image1026×394 21 KB
Hello, I am trying to make some code with pyannote/speaker-diarization-3.1 but I got some error that I cannot handle now….
\nThis is the code I made below, I only used function “speaker_diarization” this time..
\nimport pandas as pd\nfrom transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline\n\nfrom pyannote.audio import Pipeline\n\n\n\nfrom pathlib import Path\nimport os, sys\n\nffmpeg_dll_dir = Path(r""C:\\Users\\majh0\\miniconda3\\Library\\bin"") \nassert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir\nos.add_dll_directory(str(ffmpeg_dll_dir)) \n\n\nimport torch, torchcodec, platform, subprocess\nprint(""exe:"", sys.executable)\nprint(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\nsubprocess.run([""ffmpeg"", ""-version""], check=True)\nprint(""cuda torch?"",torch.cuda.is_available())\n\n\n\n\ndef whisper_stt(\n audio_file_path: str,\n output_file_path: str = ""./output.csv"",\n):\n device = ""cuda:0"" if torch.cuda.is_available() else ""cpu""\n torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32\n model_id = ""openai/whisper-large-v3-turbo""\n\n model = AutoModelForSpeechSeq2Seq.from_pretrained(\n model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True\n )\n model.to(device)\n\n processor = AutoProcessor.from_pretrained(model_id)\n\n pipe = pipeline(\n ""automatic-speech-recognition"",\n model=model,\n tokenizer=processor.tokenizer,\n feature_extractor=processor.feature_extractor,\n torch_dtype=torch_dtype,\n device=device,\n return_timestamps=True, \n chunk_length_s=10, \n stride_length_s=2, \n )\n\n result = pipe(audio_file_path)\n df = whisper_to_dataframe(result, output_file_path)\n\n return result, df\n\n\n\ndef whisper_to_dataframe(result, output_file_path):\n start_end_text = []\n\n for chunk in result[""chunks""]:\n start = chunk[""timestamp""][0]\n end = chunk[""timestamp""][1]\n text = chunk[""text""]\n start_end_text.append([start, end, text])\n df = pd.DataFrame(start_end_text, columns=[""start"", ""end"", ""text""])\n df.to_csv(output_file_path, index=False, sep=""|"")\n \n return df\n\n\ndef speaker_diarization(\n audio_file_path: str,\n output_rttm_file_path: str,\n output_csv_file_path: str,\n):\n pipeline = Pipeline.from_pretrained(\n ""pyannote/speaker-diarization-3.1"",\n token="""")\n\n if torch.cuda.is_available():\n pipeline.to(torch.device(""cuda""))\n print(""Using CUDA"")\n else:\n print(""Using CPU"")\n \n print(""torch version:"", torch.__version__)\n print(""compiled with cuda:"", torch.version.cuda)\n print(""cuda available:"", torch.cuda.is_available())\n\n out = pipeline(audio_file_path)\n ann = out.speaker_diarization\n\n # dump the diarization output to disk using RTTM format\n with open(output_rttm_file_path, ""w"", encoding=""utf-8"") as rttm:\n ann.write_rttm(rttm)\n\n df_rttm = pd.read_csv(\n output_rttm_file_path,\n sep=\' \',\n header=None,\n names=[\'type\', \'file\', \'chnl\', \'start\', \'duration\', \'C1\', \'C2\', \'speaker_id\', \'C3\', \'C4\']\n)\n \n\n df_rttm[\'end\'] = df_rttm[\'start\'] + df_rttm[\'duration\']\n\n\n df_rttm[""number""] = None\n df_rttm.at[0, ""number""] = 0\n\n\n for i in range(1, len(df_rttm)):\n if df_rttm.at[i, ""speaker_id""] != df_rttm.at[i-1, ""speaker_id""]:\n df_rttm.at[i, ""number""] = df_rttm.at[i-1, ""number""] + 1\n else:\n df_rttm.at[i, ""number""] = df_rttm.at[i-1, ""number""]\n\n\n\n df_rttm_grouped = df_rttm.groupby(""number"").agg(\n start=pd.NamedAgg(column=""start"", aggfunc=""min""),\n end=pd.NamedAgg(column=""end"", aggfunc=""max""),\n speaker_id=pd.NamedAgg(column=""speaker_id"", aggfunc=""first"")\n )\n\n df_rttm_grouped[\'duration\'] = df_rttm_grouped[\'end\'] - df_rttm_grouped[\'start\']\n df_rttm_grouped = df_rttm_grouped.reset_index(drop=True)\n\n\n df_rttm_grouped.to_csv(output_csv_file_path, sep=\',\', index=False, encoding=\'utf-8\')\n\n return df_rttm_grouped\n\n\n\n\n\nif __name__ == ""__main__"":\n # result, df = whisper_stt(\n # ""./chap05/guitar.wav"",\n # ""./chap05/guitar.csv"",\n # )\n\n # print(df)\n\n\n audio_file_path = ""./chap05/guitar.wav""\n stt_output_file_path = ""./chap05/guitar.csv""\n rttm_file_path = ""./chap05/guitar.rttm""\n rttm_csv_file_path = ""./chap05/guitar_rttm.csv""\n\n df_rttm = speaker_diarization(\n audio_file_path,\n rttm_file_path,\n rttm_csv_file_path\n )\n\n print(df_rttm)\nAfter running this code, it gives me error like below..
\n(venv) PS C:\\GPT_AGENT_2025_BOOK> & C:/GPT_AGENT_2025_BOOK/venv/Scripts/python.exe c:/GPT_AGENT_2025_BOOK/chap05/whisper_stt.py\nC:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\core\\io.py:47: UserWarning: \ntorchcodec is not installed correctly so built-in audio decoding will fail. Solutions are:\n* use audio preloaded in-memory as a {\'waveform\': (channel, time) torch.Tensor, \'sample_rate\': int} dictionary;\n* fix torchcodec installation. Error message was:\n\nCould not load libtorchcodec. Likely causes:\n 1. FFmpeg is not properly installed in your environment. We support\n versions 4, 5, 6 and 7.\n 2. The PyTorch version (2.9.0+cu126) is not compatible with\n this version of TorchCodec. Refer to the version compatibility\n table:\n https://github.com/pytorch/torchcodec?tab=readme-ov-file#installing-torchcodec.\n 3. Another runtime dependency; see exceptions below.\n The following exceptions were raised as we tried to load libtorchcodec:\n\n[start of libtorchcodec loading traceback]\nFFmpeg version 8: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core8.dll\nFFmpeg version 7: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core7.dll\nFFmpeg version 6: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core6.dll\nFFmpeg version 5: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core5.dll\nFFmpeg version 4: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core4.dll\n[end of libtorchcodec loading traceback].\n warnings.warn(\nexe: C:\\GPT_AGENT_2025_BOOK\\venv\\Scripts\\python.exe\ntorch 2.9.0+cu126 torchcodec 0.8.0 py 3.12.9\nffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers\nbuilt with gcc 10.2.1 (GCC) 20200726\nconfiguration: --disable-static --enable-shared --enable-gpl --enable-version3 --enable-sdl2 --enable-fontconfig --enable-gnutls --enable-iconv --enable-libass --enable-libdav1d --enable-libbluray --enable-libfreetype --enable-libmp3lame --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libopus --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libsrt --enable-libtheora --enable-libtwolame --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libzimg --enable-lzma --enable-zlib --enable-gmp --enable-libvidstab --enable-libvmaf --enable-libvorbis --enable-libvo-amrwbenc --enable-libmysofa --enable-libspeex --enable-libxvid --enable-libaom --enable-libgsm --enable-librav1e --disable-w32threads --enable-libmfx --enable-ffnvcodec --enable-cuda-llvm --enable-cuvid --enable-d3d11va --enable-nvenc --enable-nvdec --enable-dxva2 --enable-avisynth --enable-libopenmpt --enable-amf\nlibavutil 56. 51.100 / 56. 51.100\nlibavcodec 58. 91.100 / 58. 91.100\nlibavformat 58. 45.100 / 58. 45.100\nlibavdevice 58. 10.100 / 58. 10.100\nlibavfilter 7. 85.100 / 7. 85.100\nlibswscale 5. 7.100 / 5. 7.100\nlibswresample 3. 7.100 / 3. 7.100\nlibpostproc 55. 7.100 / 55. 7.100\ncuda torch? True\nUsing CUDA\ntorch version: 2.9.0+cu126\ncompiled with cuda: 12.6\ncuda available: True\nC:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torch\\backends\\cuda\\__init__.py:131: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = \'tf32\' \nor torch.backends.cuda.matmul.fp32_precision = \'ieee\'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at C:\\actions-runner\\_work\\pytorch\\pytorch\\pytorch\\aten\\src\\ATen\\Context.cpp:85.)\n return torch._C._get_cublas_allow_tf32()\nC:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\utils\\reproducibility.py:74: ReproducibilityWarning: TensorFloat-32 (TF32) has been disabled as it might lead to reproducibility issues and lower accuracy.\nIt can be re-enabled by calling\n >>> import torch\n >>> torch.backends.cuda.matmul.allow_tf32 = True\n >>> torch.backends.cudnn.allow_tf32 = True\nSee https://github.com/pyannote/pyannote-audio/issues/1370 for more details.\n\n warnings.warn(\nTraceback (most recent call last):\n File ""c:\\GPT_AGENT_2025_BOOK\\chap05\\whisper_stt.py"", line 156, in <module>\n df_rttm = speaker_diarization(\n ^^^^^^^^^^^^^^^^^^^^\n File ""c:\\GPT_AGENT_2025_BOOK\\chap05\\whisper_stt.py"", line 94, in speaker_diarization\n out = pipeline(audio_file_path)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File ""C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\core\\pipeline.py"", line 440, in __call__\n track_pipeline_apply(self, file, **kwargs)\n File ""C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\telemetry\\metrics.py"", line 152, in track_pipeline_apply\n duration: float = Audio().get_duration(file)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^\n File ""C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\core\\io.py"", line 273, in get_duration\n metadata: AudioStreamMetadata = get_audio_metadata(file)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File ""C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\core\\io.py"", line 86, in get_audio_metadata\n metadata = AudioDecoder(file[""audio""]).metadata\n ^^^^^^^^^^^^\nNameError: name \'AudioDecoder\' is not defined\nIt says torchcodec is not installed so auodio decoding will fail.. but strange thing is that it tells me the version of torch codec as below….
\nC:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\core\\io.py:47: UserWarning: \ntorchcodec is not installed correctly so built-in audio decoding will fail.\n\n\n(...)\n\n[end of libtorchcodec loading traceback].\n warnings.warn(\nexe: C:\\GPT_AGENT_2025_BOOK\\venv\\Scripts\\python.exe\ntorch 2.9.0+cu126 torchcodec 0.8.0 py 3.12.9\nffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers\nbuilt with gcc 10.2.1 (GCC) 20200726\nand more strange thing is that this code actually worked pretty well without any problem in Jupyternote book… and last picture is the result..
\n\n\n\nIt is hard to understand for me because I didn’t change any environment setting… and I just almost copied and pasted the code from the Jupyternote book..
\nThank you so much for the help in advance…
', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-25T07:56:14.768Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 48, 'reads': 5, 'readers_count': 4, 'score': 246.0, 'yours': False, 'topic_id': 169415, 'topic_slug': 'problem-with-pyannote-speaker-diarization-3-1', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-speaker-diarization-3-1/169415/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 244112, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-25T07:31:53.165Z', 'cooked': '', 'post_number': 2, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-10-25T07:31:53.165Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 169415, 'topic_slug': 'problem-with-pyannote-speaker-diarization-3-1', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'visible.disabled', 'post_url': '/t/problem-with-pyannote-speaker-diarization-3-1/169415/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 244126, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-25T07:56:14.176Z', 'cooked': '', 'post_number': 3, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-10-25T07:56:14.176Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 169415, 'topic_slug': 'problem-with-pyannote-speaker-diarization-3-1', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'visible.enabled', 'post_url': '/t/problem-with-pyannote-speaker-diarization-3-1/169415/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 244133, 'name': 'MAJH', 'username': 'aldkela', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/4bbf92/{size}.png', 'created_at': '2025-10-25T08:44:46.837Z', 'cooked': 'I am so sorry for this…
\nI uploaded a few threads with the same topic….
\nPlease ignore this thread..
\nI am really sorry for this inconvenience…
', 'post_number': 4, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-25T14:59:09.677Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 70.6, 'yours': False, 'topic_id': 169415, 'topic_slug': 'problem-with-pyannote-speaker-diarization-3-1', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-speaker-diarization-3-1/169415/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 244136, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-25T08:53:27.062Z', 'cooked': 'Problems frequently occur in Windows environments.
\nSpecifically, issues related to DLLs can arise because Python 3.8 and later no longer reference the Windows PATH environment variable.
Hello!
\nI just changed the code “out = pipeline(audio_file)” to the one you gave me
\nwaveform, sr = torchaudio.load(audio_file_path)\n\nout = pipeline({""waveform"": waveform, ""sample_rate"": sr})\nIt magically works!!
\nBy the way, How did you find the solution that fast? and even you made this document so fast!
\n\n\nDid you used the Chat GPT to find the solution?
\nAnyways, Thank you so much for your help again and I think you are really good at programming!
', 'post_number': 6, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-26T03:54:02.655Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 169415, 'topic_slug': 'problem-with-pyannote-speaker-diarization-3-1', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum2/blob/main/torchcodec_windows_error_1.md', 'internal': False, 'reflection': False, 'title': 'torchcodec_windows_error_1.md · John6666/forum2 at main', 'clicks': 1}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-speaker-diarization-3-1/169415/6', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 244195, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-26T04:23:33.479Z', 'cooked': '\n\nBy the way, How did you find the solution that fast? and even you made this document so fast!
\n
Yeah. Since it was an error I recognized from a similar case, I fed my prior knowledge to GPT-5 Thinking and had it search for it. I then formatted that Markdown in Python and output it.
\nI think Gemini can do it too…
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-10-26T16:23:43.476Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 169415, 'topic_slug': 'problem-with-pyannote-speaker-diarization-3-1', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/problem-with-pyannote-speaker-diarization-3-1/169415/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello, I am trying to make some code with pyannote/speaker-diarization-3.1 but I got some error that I cannot handle now….
+This is the code I made below, I only used function “speaker_diarization” this time..
+import pandas as pd
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
+
+from pyannote.audio import Pipeline
+
+
+
+from pathlib import Path
+import os, sys
+
+ffmpeg_dll_dir = Path(r""C:\Users\majh0\miniconda3\Library\bin"")
+assert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir
+os.add_dll_directory(str(ffmpeg_dll_dir))
+
+
+import torch, torchcodec, platform, subprocess
+print(""exe:"", sys.executable)
+print(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())
+subprocess.run([""ffmpeg"", ""-version""], check=True)
+print(""cuda torch?"",torch.cuda.is_available())
+
+
+
+
+def whisper_stt(
+ audio_file_path: str,
+ output_file_path: str = ""./output.csv"",
+):
+ device = ""cuda:0"" if torch.cuda.is_available() else ""cpu""
+ torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+ model_id = ""openai/whisper-large-v3-turbo""
+
+ model = AutoModelForSpeechSeq2Seq.from_pretrained(
+ model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
+ )
+ model.to(device)
+
+ processor = AutoProcessor.from_pretrained(model_id)
+
+ pipe = pipeline(
+ ""automatic-speech-recognition"",
+ model=model,
+ tokenizer=processor.tokenizer,
+ feature_extractor=processor.feature_extractor,
+ torch_dtype=torch_dtype,
+ device=device,
+ return_timestamps=True,
+ chunk_length_s=10,
+ stride_length_s=2,
+ )
+
+ result = pipe(audio_file_path)
+ df = whisper_to_dataframe(result, output_file_path)
+
+ return result, df
+
+
+
+def whisper_to_dataframe(result, output_file_path):
+ start_end_text = []
+
+ for chunk in result[""chunks""]:
+ start = chunk[""timestamp""][0]
+ end = chunk[""timestamp""][1]
+ text = chunk[""text""]
+ start_end_text.append([start, end, text])
+ df = pd.DataFrame(start_end_text, columns=[""start"", ""end"", ""text""])
+ df.to_csv(output_file_path, index=False, sep=""|"")
+
+ return df
+
+
+def speaker_diarization(
+ audio_file_path: str,
+ output_rttm_file_path: str,
+ output_csv_file_path: str,
+):
+ pipeline = Pipeline.from_pretrained(
+ ""pyannote/speaker-diarization-3.1"",
+ token="""")
+
+ if torch.cuda.is_available():
+ pipeline.to(torch.device(""cuda""))
+ print(""Using CUDA"")
+ else:
+ print(""Using CPU"")
+
+ print(""torch version:"", torch.__version__)
+ print(""compiled with cuda:"", torch.version.cuda)
+ print(""cuda available:"", torch.cuda.is_available())
+
+ out = pipeline(audio_file_path)
+ ann = out.speaker_diarization
+
+ # dump the diarization output to disk using RTTM format
+ with open(output_rttm_file_path, ""w"", encoding=""utf-8"") as rttm:
+ ann.write_rttm(rttm)
+
+ df_rttm = pd.read_csv(
+ output_rttm_file_path,
+ sep=' ',
+ header=None,
+ names=['type', 'file', 'chnl', 'start', 'duration', 'C1', 'C2', 'speaker_id', 'C3', 'C4']
+)
+
+
+ df_rttm['end'] = df_rttm['start'] + df_rttm['duration']
+
+
+ df_rttm[""number""] = None
+ df_rttm.at[0, ""number""] = 0
+
+
+ for i in range(1, len(df_rttm)):
+ if df_rttm.at[i, ""speaker_id""] != df_rttm.at[i-1, ""speaker_id""]:
+ df_rttm.at[i, ""number""] = df_rttm.at[i-1, ""number""] + 1
+ else:
+ df_rttm.at[i, ""number""] = df_rttm.at[i-1, ""number""]
+
+
+
+ df_rttm_grouped = df_rttm.groupby(""number"").agg(
+ start=pd.NamedAgg(column=""start"", aggfunc=""min""),
+ end=pd.NamedAgg(column=""end"", aggfunc=""max""),
+ speaker_id=pd.NamedAgg(column=""speaker_id"", aggfunc=""first"")
+ )
+
+ df_rttm_grouped['duration'] = df_rttm_grouped['end'] - df_rttm_grouped['start']
+ df_rttm_grouped = df_rttm_grouped.reset_index(drop=True)
+
+
+ df_rttm_grouped.to_csv(output_csv_file_path, sep=',', index=False, encoding='utf-8')
+
+ return df_rttm_grouped
+
+
+
+
+
+if __name__ == ""__main__"":
+ # result, df = whisper_stt(
+ # ""./chap05/guitar.wav"",
+ # ""./chap05/guitar.csv"",
+ # )
+
+ # print(df)
+
+
+ audio_file_path = ""./chap05/guitar.wav""
+ stt_output_file_path = ""./chap05/guitar.csv""
+ rttm_file_path = ""./chap05/guitar.rttm""
+ rttm_csv_file_path = ""./chap05/guitar_rttm.csv""
+
+ df_rttm = speaker_diarization(
+ audio_file_path,
+ rttm_file_path,
+ rttm_csv_file_path
+ )
+
+ print(df_rttm)
+After running this code, it gives me error like below..
+(venv) PS C:\GPT_AGENT_2025_BOOK> & C:/GPT_AGENT_2025_BOOK/venv/Scripts/python.exe c:/GPT_AGENT_2025_BOOK/chap05/whisper_stt.py
+C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\core\io.py:47: UserWarning:
+torchcodec is not installed correctly so built-in audio decoding will fail. Solutions are:
+* use audio preloaded in-memory as a {'waveform': (channel, time) torch.Tensor, 'sample_rate': int} dictionary;
+* fix torchcodec installation. Error message was:
+
+Could not load libtorchcodec. Likely causes:
+ 1. FFmpeg is not properly installed in your environment. We support
+ versions 4, 5, 6 and 7.
+ 2. The PyTorch version (2.9.0+cu126) is not compatible with
+ this version of TorchCodec. Refer to the version compatibility
+ table:
+ https://github.com/pytorch/torchcodec?tab=readme-ov-file#installing-torchcodec.
+ 3. Another runtime dependency; see exceptions below.
+ The following exceptions were raised as we tried to load libtorchcodec:
+
+[start of libtorchcodec loading traceback]
+FFmpeg version 8: Could not load this library: C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\torchcodec\libtorchcodec_core8.dll
+FFmpeg version 7: Could not load this library: C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\torchcodec\libtorchcodec_core7.dll
+FFmpeg version 6: Could not load this library: C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\torchcodec\libtorchcodec_core6.dll
+FFmpeg version 5: Could not load this library: C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\torchcodec\libtorchcodec_core5.dll
+FFmpeg version 4: Could not load this library: C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\torchcodec\libtorchcodec_core4.dll
+[end of libtorchcodec loading traceback].
+ warnings.warn(
+exe: C:\GPT_AGENT_2025_BOOK\venv\Scripts\python.exe
+torch 2.9.0+cu126 torchcodec 0.8.0 py 3.12.9
+ffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers
+built with gcc 10.2.1 (GCC) 20200726
+configuration: --disable-static --enable-shared --enable-gpl --enable-version3 --enable-sdl2 --enable-fontconfig --enable-gnutls --enable-iconv --enable-libass --enable-libdav1d --enable-libbluray --enable-libfreetype --enable-libmp3lame --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libopus --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libsrt --enable-libtheora --enable-libtwolame --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libzimg --enable-lzma --enable-zlib --enable-gmp --enable-libvidstab --enable-libvmaf --enable-libvorbis --enable-libvo-amrwbenc --enable-libmysofa --enable-libspeex --enable-libxvid --enable-libaom --enable-libgsm --enable-librav1e --disable-w32threads --enable-libmfx --enable-ffnvcodec --enable-cuda-llvm --enable-cuvid --enable-d3d11va --enable-nvenc --enable-nvdec --enable-dxva2 --enable-avisynth --enable-libopenmpt --enable-amf
+libavutil 56. 51.100 / 56. 51.100
+libavcodec 58. 91.100 / 58. 91.100
+libavformat 58. 45.100 / 58. 45.100
+libavdevice 58. 10.100 / 58. 10.100
+libavfilter 7. 85.100 / 7. 85.100
+libswscale 5. 7.100 / 5. 7.100
+libswresample 3. 7.100 / 3. 7.100
+libpostproc 55. 7.100 / 55. 7.100
+cuda torch? True
+Using CUDA
+torch version: 2.9.0+cu126
+compiled with cuda: 12.6
+cuda available: True
+C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\torch\backends\cuda\__init__.py:131: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = 'tf32'
+or torch.backends.cuda.matmul.fp32_precision = 'ieee'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\aten\src\ATen\Context.cpp:85.)
+ return torch._C._get_cublas_allow_tf32()
+C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\utils\reproducibility.py:74: ReproducibilityWarning: TensorFloat-32 (TF32) has been disabled as it might lead to reproducibility issues and lower accuracy.
+It can be re-enabled by calling
+ >>> import torch
+ >>> torch.backends.cuda.matmul.allow_tf32 = True
+ >>> torch.backends.cudnn.allow_tf32 = True
+See https://github.com/pyannote/pyannote-audio/issues/1370 for more details.
+
+ warnings.warn(
+Traceback (most recent call last):
+ File ""c:\GPT_AGENT_2025_BOOK\chap05\whisper_stt.py"", line 156, in <module>
+ df_rttm = speaker_diarization(
+ ^^^^^^^^^^^^^^^^^^^^
+ File ""c:\GPT_AGENT_2025_BOOK\chap05\whisper_stt.py"", line 94, in speaker_diarization
+ out = pipeline(audio_file_path)
+ ^^^^^^^^^^^^^^^^^^^^^^^^^
+ File ""C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\core\pipeline.py"", line 440, in __call__
+ track_pipeline_apply(self, file, **kwargs)
+ File ""C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\telemetry\metrics.py"", line 152, in track_pipeline_apply
+ duration: float = Audio().get_duration(file)
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^
+ File ""C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\core\io.py"", line 273, in get_duration
+ metadata: AudioStreamMetadata = get_audio_metadata(file)
+ ^^^^^^^^^^^^^^^^^^^^^^^^
+ File ""C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\core\io.py"", line 86, in get_audio_metadata
+ metadata = AudioDecoder(file[""audio""]).metadata
+ ^^^^^^^^^^^^
+NameError: name 'AudioDecoder' is not defined
+It says torchcodec is not installed so auodio decoding will fail.. but strange thing is that it tells me the version of torch codec as below….
+C:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\core\io.py:47: UserWarning:
+torchcodec is not installed correctly so built-in audio decoding will fail.
+
+
+(...)
+
+[end of libtorchcodec loading traceback].
+ warnings.warn(
+exe: C:\GPT_AGENT_2025_BOOK\venv\Scripts\python.exe
+torch 2.9.0+cu126 torchcodec 0.8.0 py 3.12.9
+ffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers
+built with gcc 10.2.1 (GCC) 20200726
+and more strange thing is that this code actually worked pretty well without any problem in Jupyternote book… and last picture is the result..
+ + + +It is hard to understand for me because I didn’t change any environment setting… and I just almost copied and pasted the code from the Jupyternote book..
+Thank you so much for the help in advance…
","Problems frequently occur in Windows environments.
+Specifically, issues related to DLLs can arise because Python 3.8 and later no longer reference the Windows PATH environment variable.
Hi everyone,
\nI’ve been trying to switch from LoRA to QLoRA on an Nvidia T4, but I’m running into an issue where the evaluation loss stays completely flat, while the training loss fluctuates around its initial value.
My LoRA setup works fine, but adding bnb_config, model.gradient_checkpointing_enable(), and model = prepare_model_for_kbit_training(model) causes the issue described above.
\n
Since the non-quantized version runs without problems, I don’t think the issue is related to the LoRA config, dataset, or formatting functions. The number of trainable parameters is non-zero for both the LoRA and QLoRA setups.
\nBelow is the code I’m using for QLoRA. Any help would be appreciated!
\nds_train_with_assistant_content = ds_train.map(construct_message_with_assistant_content)\nds_valid_with_assistant_content = ds_valid.map(construct_message_with_assistant_content)\n\nbnb_config = BitsAndBytesConfig(\n load_in_4bit=True,\n bnb_4bit_use_double_quant=True,\n bnb_4bit_quant_type=""nf4"",\n bnb_4bit_compute_dtype=torch.bfloat16\n)\n\ncheckpoint = ""Qwen/Qwen3-0.6B""\ntokenizer = AutoTokenizer.from_pretrained(checkpoint)\nmodel = AutoModelForCausalLM.from_pretrained(\n checkpoint,\n device_map=""auto"",\n quantization_config=bnb_config\n)\n\nmodel.config.use_cache = False\nmodel.gradient_checkpointing_enable()\nmodel = prepare_model_for_kbit_training(model)\nmodel.enable_input_require_grads()\n\n\ntimestamp = datetime.now().strftime(\'%Y-%m-%d_%H-%M-%S\')\nRUN_NAME = f\'qlora-final-model-all-linear-r64-{timestamp}\'\nwandb.init(\n project=os.environ[""WANDB_PROJECT""],\n name=RUN_NAME,\n # id=run_id, # resume previous run if available\n resume=""allow"", # allows resuming crashed run\n)\n\n\nRESUME_TRAINING = False\nOUTPUT_DIR = ""./qlora-final_model_all_linear_r64-output""\nPER_DEVICE_BATCH_SIZE = 2 # higher values --> OOM\n\noptimizer = \'paged_adamw_8bit\'\neffective_batch_size = 16\nlearning_rate = 1e-5\nweight_decay = 0.0\nbetas = (0.9, 0.9999)\nwarmup_ratio = 0.2\nepochs = 1\ngradient_accumulation_steps = int(effective_batch_size / PER_DEVICE_BATCH_SIZE)\nlora_r = 16*4\nlora_alpha = 64*4\nlora_dropout = 0.01\n\n\ntraining_args = TrainingArguments(\n output_dir=OUTPUT_DIR,\n per_device_train_batch_size=PER_DEVICE_BATCH_SIZE,\n gradient_accumulation_steps=gradient_accumulation_steps,\n learning_rate=learning_rate,\n optim=optimizer, \n num_train_epochs=epochs,\n weight_decay=weight_decay,\n lr_scheduler_type=""cosine"",\n warmup_ratio=warmup_ratio,\n save_strategy=""steps"",\n save_steps=gradient_accumulation_steps*5,\n save_total_limit=2,\n eval_strategy=""steps"",\n eval_steps=gradient_accumulation_steps*5,\n logging_strategy=""steps"",\n logging_steps=gradient_accumulation_steps*5,\n report_to=[\'wandb\'],\n run_name=RUN_NAME,\n bf16=True,\n # fp16=True,\n # fp16_full_eval=True,\n metric_for_best_model=""eval_loss"",\n greater_is_better=False,\n max_grad_norm=1,\n load_best_model_at_end=True,\n gradient_checkpointing=True,\n gradient_checkpointing_kwargs={""use_reentrant"": False}\n)\n\n\npeft_config = LoraConfig(\n r=lora_r,\n lora_alpha=lora_alpha,\n lora_dropout=lora_dropout,\n bias=""none"",\n task_type=""CAUSAL_LM"",\n target_modules=\'all-linear\'\n)\n# model.requires_grad_(False) # freeze base weights (precautionary)\nmodel_peft = get_peft_model(model, peft_config) # inject a LoRA adapter\nprint_trainable_parameters(model_peft)\n\ntrainer = SFTTrainer(\n model=model_peft,\n train_dataset=ds_train_with_assistant_content,\n eval_dataset=ds_valid_with_assistant_content,\n formatting_func=formatting_func,\n args=training_args,\n callbacks=[EarlyStoppingCallback(early_stopping_patience=25)]\n)\n\n\n# Training setup summary\ndataset_size = len(ds_train_with_assistant_content)\nsteps_per_epoch = dataset_size // (PER_DEVICE_BATCH_SIZE * gradient_accumulation_steps)\ntotal_steps = steps_per_epoch * epochs\nwarmup_steps = int(total_steps * warmup_ratio)\n\nprint(""===== Training Setup Summary ====="")\nprint(f""Num epochs: {epochs}"")\nprint(f""Effective batch size: {effective_batch_size}"")\nprint(f""Per-device batch size: {PER_DEVICE_BATCH_SIZE}"")\nprint(f""Gradient accumulation: {gradient_accumulation_steps}"")\nprint(f""Dataset size: {dataset_size}"")\nprint(f""Steps per epoch: {steps_per_epoch}"")\nprint(f""Total training steps: {total_steps}"")\nprint(f""Warmup steps: {warmup_steps}"")\nprint(f""Logging steps: {training_args.logging_steps}"")\nprint(""==================================="")\nprint(f""Start time: {datetime.now().strftime(\'%Y-%m-%d_%H-%M-%S\')}"")\n\n\n# Training\nlast_checkpoint = None\nif RESUME_TRAINING and os.path.isdir(OUTPUT_DIR):\n last_checkpoint = get_last_checkpoint(OUTPUT_DIR)\n\nif last_checkpoint is not None:\n print(f""Resuming training from checkpoint: {last_checkpoint}"")\n trainer.train(resume_from_checkpoint=last_checkpoint)\nelse:\n print(""Starting fresh training run"")\n trainer.train()\n\nprint(f""End time: {datetime.now().strftime(\'%Y-%m-%d_%H-%M-%S\')}"")\n\n\n# WandB logging of eval metrics\nfor log in trainer.state.log_history:\n if \'eval_loss\' in log:\n wandb.log({\n ""eval_loss"": log[\'eval_loss\'],\n ""eval_perplexity"": math.exp(log[\'eval_loss\']),\n ""step"": log[\'step\'],\n ""learning_rate"": learning_rate,\n ""weight_decay"": weight_decay,\n ""betas"": betas,\n ""warmup_ratio"": warmup_ratio,\n ""effective_batch_size"": effective_batch_size,\n ""optimizer"": optimizer\n })\n\nwandb.finish() # finish the run\n\nNvidia T4
\n
Since T4 doesn’t natively support torch.bfloat16, using torch.float16/ fp16=True instead might resolve the error. No other major issues appear to exist.
Thanks for the suggestion
\nIt turned out the issue was environment-related — I was able to get the expected results using the exact same code on Colab. In my local environment, clearing the caches for transformers, torch, etc., and upgrading all the libraries resolved the problem.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-24T18:16:57.733Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 0, 'yours': False, 'topic_id': 169337, 'topic_slug': 'qlora-model-isnt-training', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/qlora-model-isnt-training/169337/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I’ve been trying to switch from LoRA to QLoRA on an Nvidia T4, but I’m running into an issue where the evaluation loss stays completely flat, while the training loss fluctuates around its initial value.
My LoRA setup works fine, but adding bnb_config, model.gradient_checkpointing_enable(), and model = prepare_model_for_kbit_training(model) causes the issue described above.
+
Since the non-quantized version runs without problems, I don’t think the issue is related to the LoRA config, dataset, or formatting functions. The number of trainable parameters is non-zero for both the LoRA and QLoRA setups.
+Below is the code I’m using for QLoRA. Any help would be appreciated!
+ds_train_with_assistant_content = ds_train.map(construct_message_with_assistant_content)
+ds_valid_with_assistant_content = ds_valid.map(construct_message_with_assistant_content)
+
+bnb_config = BitsAndBytesConfig(
+ load_in_4bit=True,
+ bnb_4bit_use_double_quant=True,
+ bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_compute_dtype=torch.bfloat16
+)
+
+checkpoint = ""Qwen/Qwen3-0.6B""
+tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+model = AutoModelForCausalLM.from_pretrained(
+ checkpoint,
+ device_map=""auto"",
+ quantization_config=bnb_config
+)
+
+model.config.use_cache = False
+model.gradient_checkpointing_enable()
+model = prepare_model_for_kbit_training(model)
+model.enable_input_require_grads()
+
+
+timestamp = datetime.now().strftime('%Y-%m-%d_%H-%M-%S')
+RUN_NAME = f'qlora-final-model-all-linear-r64-{timestamp}'
+wandb.init(
+ project=os.environ[""WANDB_PROJECT""],
+ name=RUN_NAME,
+ # id=run_id, # resume previous run if available
+ resume=""allow"", # allows resuming crashed run
+)
+
+
+RESUME_TRAINING = False
+OUTPUT_DIR = ""./qlora-final_model_all_linear_r64-output""
+PER_DEVICE_BATCH_SIZE = 2 # higher values --> OOM
+
+optimizer = 'paged_adamw_8bit'
+effective_batch_size = 16
+learning_rate = 1e-5
+weight_decay = 0.0
+betas = (0.9, 0.9999)
+warmup_ratio = 0.2
+epochs = 1
+gradient_accumulation_steps = int(effective_batch_size / PER_DEVICE_BATCH_SIZE)
+lora_r = 16*4
+lora_alpha = 64*4
+lora_dropout = 0.01
+
+
+training_args = TrainingArguments(
+ output_dir=OUTPUT_DIR,
+ per_device_train_batch_size=PER_DEVICE_BATCH_SIZE,
+ gradient_accumulation_steps=gradient_accumulation_steps,
+ learning_rate=learning_rate,
+ optim=optimizer,
+ num_train_epochs=epochs,
+ weight_decay=weight_decay,
+ lr_scheduler_type=""cosine"",
+ warmup_ratio=warmup_ratio,
+ save_strategy=""steps"",
+ save_steps=gradient_accumulation_steps*5,
+ save_total_limit=2,
+ eval_strategy=""steps"",
+ eval_steps=gradient_accumulation_steps*5,
+ logging_strategy=""steps"",
+ logging_steps=gradient_accumulation_steps*5,
+ report_to=['wandb'],
+ run_name=RUN_NAME,
+ bf16=True,
+ # fp16=True,
+ # fp16_full_eval=True,
+ metric_for_best_model=""eval_loss"",
+ greater_is_better=False,
+ max_grad_norm=1,
+ load_best_model_at_end=True,
+ gradient_checkpointing=True,
+ gradient_checkpointing_kwargs={""use_reentrant"": False}
+)
+
+
+peft_config = LoraConfig(
+ r=lora_r,
+ lora_alpha=lora_alpha,
+ lora_dropout=lora_dropout,
+ bias=""none"",
+ task_type=""CAUSAL_LM"",
+ target_modules='all-linear'
+)
+# model.requires_grad_(False) # freeze base weights (precautionary)
+model_peft = get_peft_model(model, peft_config) # inject a LoRA adapter
+print_trainable_parameters(model_peft)
+
+trainer = SFTTrainer(
+ model=model_peft,
+ train_dataset=ds_train_with_assistant_content,
+ eval_dataset=ds_valid_with_assistant_content,
+ formatting_func=formatting_func,
+ args=training_args,
+ callbacks=[EarlyStoppingCallback(early_stopping_patience=25)]
+)
+
+
+# Training setup summary
+dataset_size = len(ds_train_with_assistant_content)
+steps_per_epoch = dataset_size // (PER_DEVICE_BATCH_SIZE * gradient_accumulation_steps)
+total_steps = steps_per_epoch * epochs
+warmup_steps = int(total_steps * warmup_ratio)
+
+print(""===== Training Setup Summary ====="")
+print(f""Num epochs: {epochs}"")
+print(f""Effective batch size: {effective_batch_size}"")
+print(f""Per-device batch size: {PER_DEVICE_BATCH_SIZE}"")
+print(f""Gradient accumulation: {gradient_accumulation_steps}"")
+print(f""Dataset size: {dataset_size}"")
+print(f""Steps per epoch: {steps_per_epoch}"")
+print(f""Total training steps: {total_steps}"")
+print(f""Warmup steps: {warmup_steps}"")
+print(f""Logging steps: {training_args.logging_steps}"")
+print(""==================================="")
+print(f""Start time: {datetime.now().strftime('%Y-%m-%d_%H-%M-%S')}"")
+
+
+# Training
+last_checkpoint = None
+if RESUME_TRAINING and os.path.isdir(OUTPUT_DIR):
+ last_checkpoint = get_last_checkpoint(OUTPUT_DIR)
+
+if last_checkpoint is not None:
+ print(f""Resuming training from checkpoint: {last_checkpoint}"")
+ trainer.train(resume_from_checkpoint=last_checkpoint)
+else:
+ print(""Starting fresh training run"")
+ trainer.train()
+
+print(f""End time: {datetime.now().strftime('%Y-%m-%d_%H-%M-%S')}"")
+
+
+# WandB logging of eval metrics
+for log in trainer.state.log_history:
+ if 'eval_loss' in log:
+ wandb.log({
+ ""eval_loss"": log['eval_loss'],
+ ""eval_perplexity"": math.exp(log['eval_loss']),
+ ""step"": log['step'],
+ ""learning_rate"": learning_rate,
+ ""weight_decay"": weight_decay,
+ ""betas"": betas,
+ ""warmup_ratio"": warmup_ratio,
+ ""effective_batch_size"": effective_batch_size,
+ ""optimizer"": optimizer
+ })
+
+wandb.finish() # finish the runThanks for the suggestion
+It turned out the issue was environment-related — I was able to get the expected results using the exact same code on Colab. In my local environment, clearing the caches for transformers, torch, etc., and upgrading all the libraries resolved the problem.
Hello, I was trying to use model named pyannote/speaker-diarization-3.1
\nso I installed some libraries as below
\n%pip install pyannote.audio==3.1.0\n%pip install numpy==1.26\nHere is the result and I think I installed this properly…
\nCollecting pyannote.audio==3.1.0\n Using cached pyannote.audio-3.1.0-py2.py3-none-any.whl.metadata (7.8 kB)\nRequirement already satisfied: asteroid-filterbanks>=0.4 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (0.4.0)\nRequirement already satisfied: einops>=0.6.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (0.8.1)\nRequirement already satisfied: huggingface-hub>=0.13.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (0.35.3)\nRequirement already satisfied: lightning>=2.0.1 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (2.5.5)\nRequirement already satisfied: omegaconf<3.0,>=2.1 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (2.3.0)\nRequirement already satisfied: pyannote.core>=5.0.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (6.0.1)\nRequirement already satisfied: pyannote.database>=5.0.1 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (6.1.0)\nRequirement already satisfied: pyannote.metrics>=3.2 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (4.0.0)\nRequirement already satisfied: pyannote.pipeline>=3.0.1 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (4.0.0)\nRequirement already satisfied: pytorch-metric-learning>=2.1.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (2.9.0)\nRequirement already satisfied: rich>=12.0.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (14.2.0)\nRequirement already satisfied: semver>=3.0.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (3.0.4)\nRequirement already satisfied: soundfile>=0.12.1 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (0.13.1)\nRequirement already satisfied: speechbrain>=0.5.14 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (1.0.3)\nRequirement already satisfied: tensorboardX>=2.6 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (2.6.4)\nRequirement already satisfied: torch>=2.0.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (2.9.0+cu126)\nRequirement already satisfied: torch-audiomentations>=0.11.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (0.12.0)\nRequirement already satisfied: torchaudio>=2.0.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (2.9.0)\nRequirement already satisfied: torchmetrics>=0.11.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from pyannote.audio==3.1.0) (1.8.2)\nRequirement already satisfied: antlr4-python3-runtime==4.9.* in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from omegaconf<3.0,>=2.1->pyannote.audio==3.1.0) (4.9.3)\nRequirement already satisfied: PyYAML>=5.1.0 in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from omegaconf<3.0,>=2.1->pyannote.audio==3.1.0) (6.0.3)\nRequirement already satisfied: numpy in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from asteroid-filterbanks>=0.4->pyannote.audio==3.1.0) (1.26.0)\nRequirement already satisfied: typing-extensions in c:\\gpt_agent_2025_book\\venv\\lib\\site-packages (from asteroid-filterbanks>=0.4->pyannote.audio==3.1.0) (4.15.0)\n...\n Uninstalling numpy-2.3.4:\n Successfully uninstalled numpy-2.3.4\nSuccessfully installed numpy-1.26.0\nNote: you may need to restart the kernel to use updated packages.\nOutput is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...\nERROR: pip\'s dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\npyannote-core 6.0.1 requires numpy>=2.0, but you have numpy 1.26.0 which is incompatible.\npyannote-metrics 4.0.0 requires numpy>=2.2.2, but you have numpy 1.26.0 which is incompatible.\nI ran this code to load the ffmpeg
\nfrom pathlib import Path\nimport os, sys\n\nffmpeg_dll_dir = Path(r""C:\\Users\\majh0\\miniconda3\\Library\\bin"") \nassert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir\nos.add_dll_directory(str(ffmpeg_dll_dir)) \n\nimport torch, torchcodec, platform, subprocess\nprint(""exe:"", sys.executable)\nprint(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\nsubprocess.run([""ffmpeg"", ""-version""], check=True)\nprint(""cuda torch?"",torch.cuda.is_available())\nand the result looks fine to me..
\nexe: c:\\GPT_AGENT_2025_BOOK\\venv\\Scripts\\python.exe\ntorch 2.9.0+cu126 torchcodec 0.8.0 py 3.12.9\ncuda torch? True\nI ran this code and it gave me an error as below…
\n# instantiate the pipeline\nimport torch\nfrom pyannote.audio import Pipeline\npipeline = Pipeline.from_pretrained(\n ""pyannote/speaker-diarization-3.1"",\n token=""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"")\n\n\nif torch.cuda.is_available():\n pipeline.to(torch.device(""cuda""))\n print(""Using CUDA"")\nelse:\n print(""Using CPU"")\n---------------------------------------------------------------------------\nAttributeError Traceback (most recent call last)\nCell In[3], line 3\n 1 # instantiate the pipeline\n 2 import torch\n----> 3 from pyannote.audio import Pipeline\n 4 pipeline = Pipeline.from_pretrained(\n 5 ""pyannote/speaker-diarization-3.1"",\n 6 token=""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"")\n 9 if torch.cuda.is_available():\n\nFile c:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\__init__.py:29\n 25 except ImportError:\n 26 pass\n---> 29 from .core.inference import Inference\n 30 from .core.io import Audio\n 31 from .core.model import Model\n\nFile c:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\pyannote\\audio\\core\\inference.py:36\n 33 from pyannote.core import Segment, SlidingWindow, SlidingWindowFeature\n 34 from pytorch_lightning.utilities.memory import is_oom_error\n---> 36 from pyannote.audio.core.io import AudioFile\n 37 from pyannote.audio.core.model import Model, Specifications\n 38 from pyannote.audio.core.task import Resolution\n...\n 49 - a ""str"" or ""Path"" instance: ""audio.wav"" or Path(""audio.wav"")\n (...) 56 integer to load a specific channel: {""audio"": ""stereo.wav"", ""channel"": 0}\n 57 """"""\n\nAttributeError: module \'torchaudio\' has no attribute \'set_audio_backend\'\nI have checked the document and it says I need to install pyannote.audio 3.1
I don’t know why this thing doesn’t work…. I tried to solve this problem for 3hrs changing version of pyannote.audio but this thing didn’t give me solution..
\nDo I need to delete venv and reinstall it clearly..?
\nThank you so much for the help in advance..
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-10-21T14:42:42.475Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 84, 'reads': 5, 'readers_count': 4, 'score': 221.0, 'yours': False, 'topic_id': 169326, 'topic_slug': 'problem-with-pyannote-audio-3-1-0', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/pyannote/pyannote-audio', 'internal': False, 'reflection': False, 'title': 'GitHub - pyannote/pyannote-audio: Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-audio-3-1-0/169326/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243939, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-22T02:49:32.789Z', 'cooked': 'Seems library version incompatibility…
\nYour import error comes from an API removal in torchaudio and an incompatible NumPy pin. Fix by upgrading pyannote.audio and undoing the NumPy downgrade. Keep your Torch 2.9 stack.
# clean conflicting pins\npip uninstall -y pyannote.audio pyannote.core pyannote.metrics pyannote.pipeline pyannote.database numpy\n\n# install a compatible, modern set\npip install --upgrade ""numpy>=2.3"" ""pyannote.audio>=4.0.1"" --prefer-binary\n# keep your existing torch==2.9.*, torchaudio==2.9.* and torchcodec\npyannote.audio>=4 removed the old torchaudio backend call and uses FFmpeg via torchcodec, so the import works on torchaudio≥2.2. NumPy≥2.x satisfies pyannote-core and pyannote-metrics. (GitHub)
Then restart the kernel once. Verify:
\n# refs:\n# - torchaudio dispatcher notes: https://docs.pytorch.org/audio/main/torchaudio.html\n# - pyannote model card: https://huggingface.co/pyannote/speaker-diarization-3.1\nimport torchaudio, torchcodec\nprint(""backends:"", torchaudio.list_audio_backends()) # should show \'ffmpeg\' and/or \'soundfile\'\nfrom pyannote.audio import Pipeline\npipe = Pipeline.from_pretrained(""pyannote/speaker-diarization-3.1"", token=""hf_xxx"") # do not hardcode secrets\nset_audio_backend was deprecated, then removed in torchaudio 2.2+, which is why pyannote.audio==3.1.0 fails to import on your current torchaudio. (PyTorch Docs)
pyannote.audio==3.1.0 calls torchaudio.set_audio_backend(""soundfile""). That function is gone in torchaudio≥2.2, so import raises AttributeError. Upgrading pyannote fixes it because 4.x removed that path. (GitHub)numpy==1.26. Current pyannote ecosystem components require NumPy≥2.0 (core) and ≥2.2.2 (metrics). Pip warned correctly. Use NumPy≥2.3. (GitHub)pyannote.audio==3.1.0 (not recommended)Pick one, not both:
\n# Legacy stack that still has set_audio_backend\npip install ""torch<=2.1.2"" ""torchaudio<=2.1.2"" ""numpy>=2.0,<3"" ""pyannote.audio==3.1.0""\nor a temporary shim:
\n# WARNING: local hack to import 3.1.0 with new torchaudio\nimport torchaudio\nif not hasattr(torchaudio, ""set_audio_backend""):\n torchaudio.set_audio_backend = lambda *a, **k: None\n torchaudio.get_audio_backend = lambda: ""soundfile""\nfrom pyannote.audio import Pipeline\nThe first aligns versions to when the API existed. The second bypasses the call so you can upgrade later. (PyTorch Docs)
\npyannote/speaker-diarization-3.1 on Hugging Face and pass a valid token, or downloads will fail. (Hugging Face)pyannote.audio>=4 expects FFmpeg via torchcodec. You already verified FFmpeg and torchcodec, which matches the 4.x I/O design. (GitHub)# refs in comments:\n# https://huggingface.co/pyannote/speaker-diarization-3.1\n# https://docs.pytorch.org/audio/main/torchaudio.html\nimport torch\nfrom pyannote.audio import Pipeline\npipe = Pipeline.from_pretrained(""pyannote/speaker-diarization-3.1"", token=""hf_xxx"")\nif torch.cuda.is_available():\n pipe.to(""cuda"")\nresult = pipe(""sample.wav"") # 16 kHz mono recommended\nprint(result)\nThe model card confirms “pyannote.audio version 3.1 or higher,” so using 4.x is valid and simpler on modern Torch. (Hugging Face)
\nset_audio_backend and switched to a dispatcher. That is the precise cause of your AttributeError. (PyTorch Docs)sox/soundfile backends; use FFmpeg or in-memory audio. Explains why 4.x works on Windows with torchcodec. (GitHub)Deleting the venv is optional. Uninstall→reinstall with the versions above and one kernel restart is sufficient.
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-10-22T02:50:15.452Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 169326, 'topic_slug': 'problem-with-pyannote-audio-3-1-0', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/pyannote/pyannote-audio/releases', 'internal': False, 'reflection': False, 'title': 'Releases · pyannote/pyannote-audio · GitHub', 'clicks': 1}, {'url': 'https://github.com/pyannote/pyannote-audio/issues/1576', 'internal': False, 'reflection': False, 'title': 'Removing torchaudio.set_audio_backend(""soundfile"") · Issue #1576 · pyannote/pyannote-audio · GitHub', 'clicks': 1}, {'url': 'https://github.com/huggingface/transformers/issues/41230', 'internal': False, 'reflection': False, 'title': 'Consider forking and maintaining pyctcdecode or switch to torchaudio.models.decoder · Issue #41230 · huggingface/transformers · GitHub', 'clicks': 0}, {'url': 'https://huggingface.co/pyannote/speaker-diarization-3.1', 'internal': False, 'reflection': False, 'title': 'pyannote/speaker-diarization-3.1 · Hugging Face', 'clicks': 0}, {'url': 'https://docs.pytorch.org/audio/main/torchaudio.html', 'internal': False, 'reflection': False, 'title': 'torchaudio — Torchaudio 2.8.0 documentation', 'clicks': 0}, {'url': 'https://huggingface.co/collinbarnwell/pyannote-speaker-diarization-31', 'internal': False, 'reflection': False, 'title': 'collinbarnwell/pyannote-speaker-diarization-31 · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-audio-3-1-0/169326/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243955, 'name': 'MAJH', 'username': 'aldkela', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/4bbf92/{size}.png', 'created_at': '2025-10-22T12:34:52.198Z', 'cooked': 'Hello! Thank you so much!! I realized.. I should read the error msg properly to solve the problem!!! xD
\nI have one more problem….
\nI made a code as below..
\nfrom pathlib import Path\nimport os, sys\n\nffmpeg_dll_dir = Path(r""C:\\Users\\majh0\\miniconda3\\Library\\bin"") \nassert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir\nos.add_dll_directory(str(ffmpeg_dll_dir)) \n\nimport torch, torchcodec, platform, subprocess\nprint(""exe:"", sys.executable)\nprint(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\nsubprocess.run([""ffmpeg"", ""-version""], check=True)\nprint(""cuda torch?"",torch.cuda.is_available())\n\n# instantiate the pipeline\nimport torch\nfrom pyannote.audio import Pipeline\n\npipeline = Pipeline.from_pretrained(\n ""pyannote/speaker-diarization-3.1"",\n token=""my token"")\n\n\nif torch.cuda.is_available():\n pipeline.to(torch.device(""cuda""))\n print(""Using CUDA"")\nelse:\n print(""Using CPU"")\n\naudio_file =""./guitar.wav""\ndiarization = pipeline(audio_file)\n\n# dump the diarization output to disk using RTTM format\nwith open(""./guitar.rttm"", ""w"", encoding=""utf-8"") as rttm:\n diarization.write_rttm(rttm)\nthis thing gave me error as below…
\n---------------------------------------------------------------------------\nAttributeError Traceback (most recent call last)\nCell In[15], line 6\n 4 # dump the diarization output to disk using RTTM format\n 5 with open(""./guitar.rttm"", ""w"", encoding=""utf-8"") as rttm:\n----> 6 diarization.write_rttm(rttm)\n\nAttributeError: \'DiarizeOutput\' object has no attribute \'write_rttm\'\nThis thing is hard to understand for me… because I literally typed “diarization.write_rttm(rttm)” same with the example of this document like picture below https://huggingface.co/pyannote/speaker-diarization-3.1
\n\nthe name of the function “write_rttm” has changed? then is there any way to check the new name of it..?
\nor did I make another mistake again..?
\nI think I am bothering you too much.. but thank you so much for your help..
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-10-22T12:34:52.198Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 169326, 'topic_slug': 'problem-with-pyannote-audio-3-1-0', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/pyannote/speaker-diarization-3.1', 'internal': False, 'reflection': False, 'title': 'pyannote/speaker-diarization-3.1 · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-audio-3-1-0/169326/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243956, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-22T12:48:54.185Z', 'cooked': 'It seems like a partial hit. The cause is a specification change due to a library version upgrade, but it appears to be because the returned object changed, not because the function itself changed.
You’re on pyannote.audio 4.x. In 4.x the pipeline returns a DiarizeOutput object, not an Annotation. The Annotation lives at output.speaker_diarization. write_rttm is a method of Annotation, so call it there.
from pyannote.audio import Pipeline\nimport torch\n\npipeline = Pipeline.from_pretrained(\n ""pyannote/speaker-diarization-3.1"",\n token=""YOUR_HF_TOKEN""\n)\nif torch.cuda.is_available():\n pipeline.to(""cuda"")\n\nout = pipeline(""./guitar.wav"") # out is DiarizeOutput\nann = out.speaker_diarization # this is an Annotation\n\nwith open(""./guitar.rttm"", ""w"", encoding=""utf-8"") as f:\n ann.write_rttm(f)\nEvidence
\noutput = pipeline(...); for turn, spk in output.speaker_diarization: ..., proving the wrapper return type in 4.x. (GitHub)write_rttm is defined on pyannote.core.Annotation, not on the wrapper. (pyannote.github.io)Annotation directly. That is why your call failed on 4.x. (Hugging Face)Option if you want the old behavior: pin to the legacy stack (pyannote.audio==3.1.x) where pipeline(...) returns an Annotation, and the snippet diarization.write_rttm(...) works as-is. Note 4.x introduced several breaking changes, including API renames. (GitHub)
Hello, finally it works!!!
\nI thought I made mistake again.. I didn’t even think there was a change due to a library version upgrade..
\nThank you so much now I can use this model without any problem!!!
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-10-23T18:31:44.078Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 20.4, 'yours': False, 'topic_id': 169326, 'topic_slug': 'problem-with-pyannote-audio-3-1-0', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/problem-with-pyannote-audio-3-1-0/169326/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 244046, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-24T06:32:17.200Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-10-24T06:32:17.200Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 169326, 'topic_slug': 'problem-with-pyannote-audio-3-1-0', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/problem-with-pyannote-audio-3-1-0/169326/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello, I was trying to use model named pyannote/speaker-diarization-3.1
+so I installed some libraries as below
+%pip install pyannote.audio==3.1.0
+%pip install numpy==1.26
+Here is the result and I think I installed this properly…
+Collecting pyannote.audio==3.1.0
+ Using cached pyannote.audio-3.1.0-py2.py3-none-any.whl.metadata (7.8 kB)
+Requirement already satisfied: asteroid-filterbanks>=0.4 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (0.4.0)
+Requirement already satisfied: einops>=0.6.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (0.8.1)
+Requirement already satisfied: huggingface-hub>=0.13.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (0.35.3)
+Requirement already satisfied: lightning>=2.0.1 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (2.5.5)
+Requirement already satisfied: omegaconf<3.0,>=2.1 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (2.3.0)
+Requirement already satisfied: pyannote.core>=5.0.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (6.0.1)
+Requirement already satisfied: pyannote.database>=5.0.1 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (6.1.0)
+Requirement already satisfied: pyannote.metrics>=3.2 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (4.0.0)
+Requirement already satisfied: pyannote.pipeline>=3.0.1 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (4.0.0)
+Requirement already satisfied: pytorch-metric-learning>=2.1.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (2.9.0)
+Requirement already satisfied: rich>=12.0.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (14.2.0)
+Requirement already satisfied: semver>=3.0.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (3.0.4)
+Requirement already satisfied: soundfile>=0.12.1 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (0.13.1)
+Requirement already satisfied: speechbrain>=0.5.14 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (1.0.3)
+Requirement already satisfied: tensorboardX>=2.6 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (2.6.4)
+Requirement already satisfied: torch>=2.0.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (2.9.0+cu126)
+Requirement already satisfied: torch-audiomentations>=0.11.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (0.12.0)
+Requirement already satisfied: torchaudio>=2.0.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (2.9.0)
+Requirement already satisfied: torchmetrics>=0.11.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from pyannote.audio==3.1.0) (1.8.2)
+Requirement already satisfied: antlr4-python3-runtime==4.9.* in c:\gpt_agent_2025_book\venv\lib\site-packages (from omegaconf<3.0,>=2.1->pyannote.audio==3.1.0) (4.9.3)
+Requirement already satisfied: PyYAML>=5.1.0 in c:\gpt_agent_2025_book\venv\lib\site-packages (from omegaconf<3.0,>=2.1->pyannote.audio==3.1.0) (6.0.3)
+Requirement already satisfied: numpy in c:\gpt_agent_2025_book\venv\lib\site-packages (from asteroid-filterbanks>=0.4->pyannote.audio==3.1.0) (1.26.0)
+Requirement already satisfied: typing-extensions in c:\gpt_agent_2025_book\venv\lib\site-packages (from asteroid-filterbanks>=0.4->pyannote.audio==3.1.0) (4.15.0)
+...
+ Uninstalling numpy-2.3.4:
+ Successfully uninstalled numpy-2.3.4
+Successfully installed numpy-1.26.0
+Note: you may need to restart the kernel to use updated packages.
+Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...
+ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
+pyannote-core 6.0.1 requires numpy>=2.0, but you have numpy 1.26.0 which is incompatible.
+pyannote-metrics 4.0.0 requires numpy>=2.2.2, but you have numpy 1.26.0 which is incompatible.
+I ran this code to load the ffmpeg
+from pathlib import Path
+import os, sys
+
+ffmpeg_dll_dir = Path(r""C:\Users\majh0\miniconda3\Library\bin"")
+assert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir
+os.add_dll_directory(str(ffmpeg_dll_dir))
+
+import torch, torchcodec, platform, subprocess
+print(""exe:"", sys.executable)
+print(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())
+subprocess.run([""ffmpeg"", ""-version""], check=True)
+print(""cuda torch?"",torch.cuda.is_available())
+and the result looks fine to me..
+exe: c:\GPT_AGENT_2025_BOOK\venv\Scripts\python.exe
+torch 2.9.0+cu126 torchcodec 0.8.0 py 3.12.9
+cuda torch? True
+I ran this code and it gave me an error as below…
+# instantiate the pipeline
+import torch
+from pyannote.audio import Pipeline
+pipeline = Pipeline.from_pretrained(
+ ""pyannote/speaker-diarization-3.1"",
+ token=""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"")
+
+
+if torch.cuda.is_available():
+ pipeline.to(torch.device(""cuda""))
+ print(""Using CUDA"")
+else:
+ print(""Using CPU"")
+---------------------------------------------------------------------------
+AttributeError Traceback (most recent call last)
+Cell In[3], line 3
+ 1 # instantiate the pipeline
+ 2 import torch
+----> 3 from pyannote.audio import Pipeline
+ 4 pipeline = Pipeline.from_pretrained(
+ 5 ""pyannote/speaker-diarization-3.1"",
+ 6 token=""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"")
+ 9 if torch.cuda.is_available():
+
+File c:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\__init__.py:29
+ 25 except ImportError:
+ 26 pass
+---> 29 from .core.inference import Inference
+ 30 from .core.io import Audio
+ 31 from .core.model import Model
+
+File c:\GPT_AGENT_2025_BOOK\venv\Lib\site-packages\pyannote\audio\core\inference.py:36
+ 33 from pyannote.core import Segment, SlidingWindow, SlidingWindowFeature
+ 34 from pytorch_lightning.utilities.memory import is_oom_error
+---> 36 from pyannote.audio.core.io import AudioFile
+ 37 from pyannote.audio.core.model import Model, Specifications
+ 38 from pyannote.audio.core.task import Resolution
+...
+ 49 - a ""str"" or ""Path"" instance: ""audio.wav"" or Path(""audio.wav"")
+ (...) 56 integer to load a specific channel: {""audio"": ""stereo.wav"", ""channel"": 0}
+ 57 """"""
+
+AttributeError: module 'torchaudio' has no attribute 'set_audio_backend'
+I have checked the document and it says I need to install pyannote.audio 3.1
I don’t know why this thing doesn’t work…. I tried to solve this problem for 3hrs changing version of pyannote.audio but this thing didn’t give me solution..
+Do I need to delete venv and reinstall it clearly..?
+Thank you so much for the help in advance..
","It seems like a partial hit. The cause is a specification change due to a library version upgrade, but it appears to be because the returned object changed, not because the function itself changed.
You’re on pyannote.audio 4.x. In 4.x the pipeline returns a DiarizeOutput object, not an Annotation. The Annotation lives at output.speaker_diarization. write_rttm is a method of Annotation, so call it there.
from pyannote.audio import Pipeline
+import torch
+
+pipeline = Pipeline.from_pretrained(
+ ""pyannote/speaker-diarization-3.1"",
+ token=""YOUR_HF_TOKEN""
+)
+if torch.cuda.is_available():
+ pipeline.to(""cuda"")
+
+out = pipeline(""./guitar.wav"") # out is DiarizeOutput
+ann = out.speaker_diarization # this is an Annotation
+
+with open(""./guitar.rttm"", ""w"", encoding=""utf-8"") as f:
+ ann.write_rttm(f)
+Evidence
+output = pipeline(...); for turn, spk in output.speaker_diarization: ..., proving the wrapper return type in 4.x. (GitHub)write_rttm is defined on pyannote.core.Annotation, not on the wrapper. (pyannote.github.io)Annotation directly. That is why your call failed on 4.x. (Hugging Face)Option if you want the old behavior: pin to the legacy stack (pyannote.audio==3.1.x) where pipeline(...) returns an Annotation, and the snippet diarization.write_rttm(...) works as-is. Note 4.x introduced several breaking changes, including API renames. (GitHub)
Hi community,
\nHere is my image-to-text pipeline:
\n(customized means not a registered one in official Transformers)
\nA customized Image processor,
\nA VisionEncoderDecoder, with a customized vision encoder that inherits the PretrainedModel and a MBartDecoder,
\nA WordLevel tokenizer (yes I haven’t used a MBartTokenizer and I have distilled my own one for specific corpus).
\nI want to consume this pipeline in Transformers.js, however I notice that all examples given in Transformers.js documentation seem like pulling from a ready made Transformers pipeline with official components and configurations, I just wonder is it possible to turn my customized pipeline consumable for Transformers.js, or to what extent my pipeline could be partially turned to?
\nMy guess is that the I should make my own image preprocessing step and send the image input tensor to the model, in that way, which kind of js libraries you recommend to use? (It won’t be very intensive, just simply resize and normalize things plus a crop-white-margin function which doesn’t exist in Transformers’ image processors).
\nAlso just to be sure, is my VisionEncoderDecoder possible to export to an onnx format to be consumable for Transformers.js?
\nOf course my model should be possible to run in browser (and that’s the whole point for me to do this), as it has only 20M parameters (way less than the showcase in Transformers.js)
\nThanks for your help in advance!
', 'post_number': 1, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-10-08T15:19:25.343Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 26, 'reads': 9, 'readers_count': 8, 'score': 21.6, 'yours': False, 'topic_id': 169036, 'topic_slug': 'how-to-make-my-customized-pipeline-consumable-for-transformers-js', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/load-model-from-platform-other-than-hf-hub-and-display-a-progress-bar-by-from-pretrained-in-transformers-js/169364', 'internal': True, 'reflection': True, 'title': 'Load model from platform other than HF Hub and display a progress bar by `from_pretrained()` in Transformers.js', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-make-my-customized-pipeline-consumable-for-transformers-js/169036/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243331, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-08T23:15:26.000Z', 'cooked': 'It seems possible. For Transoformers.js, there’s a dedicated channel on the HF Discord, so asking there would be the most reliable option.
', 'post_number': 2, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-10-08T23:15:26.000Z', 'reply_count': 2, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 26.4, 'yours': False, 'topic_id': 169036, 'topic_slug': 'how-to-make-my-customized-pipeline-consumable-for-transformers-js', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/transformer_js_custom_pipeline_1.md', 'internal': False, 'reflection': False, 'title': 'transformer_js_custom_pipeline_1.md · John6666/forum1 at main', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-make-my-customized-pipeline-consumable-for-transformers-js/169036/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243351, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-10-09T05:47:31.103Z', 'cooked': 'Thanks let me check!
', 'post_number': 3, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-10-09T05:47:31.103Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 16.4, 'yours': False, 'topic_id': 169036, 'topic_slug': 'how-to-make-my-customized-pipeline-consumable-for-transformers-js', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-make-my-customized-pipeline-consumable-for-transformers-js/169036/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243504, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-10-13T17:27:00.991Z', 'cooked': 'Hi John,
\nI try to follow your export script and I made to export 1 onnx file with the following:
register_tasks_manager_onnx = TasksManager.create_register(""onnx"")\n@register_tasks_manager_onnx(""my_hgnetv2"", *[""feature-extraction""])\nclass HGNetv2OnnxConfig(ViTOnnxConfig):\n @property\n def inputs(self):\n return {""pixel_values"": {0: ""batch""}} # only dynamical axis is needed to list here\n @property\n def outputs(self):\n return {""last_hidden_state"": {0: ""batch""}}\n\ndef export_onnx():\n path=\'./model\'\n model = VisionEncoderDecoderModel.from_pretrained(path)\n onnx_config_constructor = TasksManager.get_exporter_config_constructor(\n exporter=""onnx"",\n model=model,\n task=""image-to-text"",\n library_name=""transformers"",\n exporter_config_kwargs={""use_past"": True},\n )\n onnx_config = onnx_config_constructor(model.config)\n out = Path(""./model/onnx"")\n out.mkdir(exist_ok=True)\n\n inputs, outputs = export(model, \n onnx_config, \n out/""model.onnx"", \n onnx_config.DEFAULT_ONNX_OPSET,\n input_shapes={""pixel_values"": [1, 3, 384, 384]},\n )\n print(inputs)\n print(outputs)\nHowever, I don’t know how to export to trio .onnx file with the cli, since within the python script, I can register the customized config, but I don’t know how to register it with cli…
', 'post_number': 4, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-10-13T17:27:47.078Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 21.2, 'yours': False, 'topic_id': 169036, 'topic_slug': 'how-to-make-my-customized-pipeline-consumable-for-transformers-js', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-make-my-customized-pipeline-consumable-for-transformers-js/169036/4', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243505, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-10-13T17:54:45.869Z', 'cooked': 'Oh I see, it’s here Export a model to ONNX with optimum.exporters.onnx and we need to use main_export instead of export
Finally I use the following:
\ndef export_onnx():\n path=\'./model\'\n out = Path(""./model/trio_onnx"")\n out.mkdir(exist_ok=True)\n\n main_export(\n path,\n task=""image-to-text"",\n output=out,\n )\nHowever, this can only export to encoder_model.onnx and decoder_model.onnx, since I have no idea how the use_past=True can be injected with main_export’s argument(The example in the above link doesn’t work out), I monkey-patched the source code to make it export to trio onnx.
For Transformer.js:
\nUse main_export() with custom_onnx_configs and with_behavior(..., use_past=True) to get the trio. Do not monkey-patch.
main_export() and choose behaviors per subgraph: ""encoder"", ""decoder"", and ""decoder with past"". You can also disable post-processing so files are kept separate. (Hugging Face)onnx/{encoder_model.onnx, decoder_model.onnx, decoder_with_past_model.onnx} or a merged decoder. (Hugging Face)# refs:\n# - Export guide (custom_onnx_configs + with_behavior + no_post_process):\n# https://huggingface.co/docs/optimum-onnx/onnx/usage_guides/export_a_model\n# - main_export reference:\n# https://huggingface.co/docs/optimum-onnx/en/onnx/package_reference/export\n\nfrom pathlib import Path\nfrom transformers import AutoConfig\nfrom optimum.exporters.onnx import main_export\nfrom optimum.exporters.tasks import TasksManager\n\nmodel_dir = ""./model"" # your VisionEncoderDecoder checkpoint\nout = Path(""./model/trio_onnx""); out.mkdir(parents=True, exist_ok=True)\n\n# Build an ONNX config for your model+task\ncfg = AutoConfig.from_pretrained(model_dir)\nctor = TasksManager.get_exporter_config_constructor(\n model_type=cfg.model_type, backend=""onnx"", task=""image-to-text"" # vision→text task\n)\nonnx_cfg = ctor(config=cfg, task=""image-to-text"")\n\n# Ask explicitly for the three subgraphs\ncustom_onnx_configs = {\n ""encoder_model"": onnx_cfg.with_behavior(""encoder""),\n ""decoder_model"": onnx_cfg.with_behavior(""decoder"", use_past=False),\n ""decoder_with_past_model"": onnx_cfg.with_behavior(""decoder"", use_past=True),\n}\n\n# Export. Keep trio separate (avoid automatic merge).\nmain_export(\n model=model_dir,\n task=""image-to-text"",\n output=str(out),\n custom_onnx_configs=custom_onnx_configs,\n no_post_process=True,\n)\nWhy this works: Optimum documents custom_onnx_configs and with_behavior(""decoder"", use_past=True) to emit decoder_with_past_model.onnx; no_post_process=True prevents the exporter from merging decoders. (Hugging Face)
encoder_model.onnx, decoder_model.onnx, decoder_with_past_model.onnx. This mirrors working web repos. (Hugging Face)custom_onnx_configs dict as above. (Hugging Face)no_post_process=True. The doc names this exact flag. (Hugging Face)TasksManager.get_supported_tasks_for_model_type(model_type, ""onnx"") and pick the vision→text task. The export guide shows this workflow. (Hugging Face)Some exporters can produce a single decoder_model_merged.onnx that handles both first and subsequent tokens. If you prefer that, omit no_post_process=True. The public ViT-GPT2 repo shows merged and split variants side by side. (Hugging Face)
Well, I still cannot make this work, by debugging, I find that the main_export() will take me to optimum.exporters.utils._get_submodels_and_export_configs(), and an error raises here
# When specifying custom export configs for supported transformers architectures, we do\n # not force to specify a custom export config for each submodel.\n for key, custom_export_config in custom_export_configs.items():\n models_and_export_configs[key] = (models_and_export_configs[key][0], custom_export_config)\nwhere the custom_export_configs is the one we passed in with use_past injected, while the models_and_export_configs, generated here
# TODO: this succession of if/else strongly suggests a refactor is needed.\n if (\n task.startswith(TasksManager._ENCODER_DECODER_TASKS)\n and model.config.is_encoder_decoder\n and not monolith\n ):\n models_and_export_configs = get_encoder_decoder_models_for_export(model, export_config)\ndoesn’t contain the key “decoder_with_past”, where the default export_config generated here
export_config_constructor = TasksManager.get_exporter_config_constructor(\n model=model, exporter=exporter, task=task, library_name=library_name\n )\n export_config = export_config_constructor(\n model.config,\n int_dtype=int_dtype,\n float_dtype=float_dtype,\n preprocessors=preprocessors,\n )\nwith a default use_past=False, therefore would not generate a config for “decoder_with_past”.
\nAnd actually here is what I monkey_patched during the debugging.
I think there is a high dependency between the export config and model config in optimum library, where I although use a customized encoder but still the VisionEncoderDecoder Config as the outermost config, which leads me to the not custom_architecture config processing logic here, which leads to the above error, which may not considered as a normal scenario in design.
if not custom_architecture:\n if library_name == ""diffusers"":\n export_config = None\n models_and_export_configs = get_diffusion_models_for_export(\n model, int_dtype=int_dtype, float_dtype=float_dtype, exporter=exporter\n )\n else:\n export_config_constructor = TasksManager.get_exporter_config_constructor(\n model=model, exporter=exporter, task=task, library_name=library_name\n )\n export_config = export_config_constructor(\n model.config,\n int_dtype=int_dtype,\n float_dtype=float_dtype,\n preprocessors=preprocessors,\n )\n\n export_config.variant = _variant\n all_variants = ""\\n"".join(\n [f"" - {name}: {description}"" for name, description in export_config.VARIANTS.items()]\n )\n logger.info(f""Using the export variant {export_config.variant}. Available variants are:\\n{all_variants}"")\n\n # TODO: this succession of if/else strongly suggests a refactor is needed.\n if (\n task.startswith(TasksManager._ENCODER_DECODER_TASKS)\n and model.config.is_encoder_decoder\n and not monolith\n ):\n models_and_export_configs = get_encoder_decoder_models_for_export(model, export_config)\n elif task.startswith(""text-generation"") and not monolith:\n models_and_export_configs = get_decoder_models_for_export(model, export_config)\n elif model.config.model_type == ""sam"":\n models_and_export_configs = get_sam_models_for_export(model, export_config)\n elif model.config.model_type == ""speecht5"":\n models_and_export_configs = get_speecht5_models_for_export(model, export_config, model_kwargs)\n elif model.config.model_type == ""musicgen"":\n models_and_export_configs = get_musicgen_models_for_export(model, export_config)\n else:\n models_and_export_configs = {""model"": (model, export_config)}\n\n # When specifying custom export configs for supported transformers architectures, we do\n # not force to specify a custom export config for each submodel.\n for key, custom_export_config in custom_export_configs.items():\n models_and_export_configs[key] = (models_and_export_configs[key][0], custom_export_config)\nAlright, actually we don’t need those verbose configs, just change the task from “image-to-text” to “image-to-text-with-past” will solve the issue (no monkey-patch)
\ndef export_onnx():\n path=\'./model\'\n out = Path(""./model/trio_onnx"")\n out.mkdir(exist_ok=True)\n main_export(\n path,\n task=""image-to-text-with-past"", # to get trio onnx model, use ""-with-past"", otherwise use ""image-to-text""\n output=out,\n )\nGreat. About _with_past
Hi John,
\nI’ve finally succeeded in implementing the above things. Thanks for your help!
\nYet I still have some other questions and I think I’d better create a new discussion.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 12, 'post_type': 3, 'posts_count': 12, 'updated_at': '2025-10-23T21:34:35.488Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 169036, 'topic_slug': 'how-to-make-my-customized-pipeline-consumable-for-transformers-js', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-make-my-customized-pipeline-consumable-for-transformers-js/169036/12', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi community,
+Here is my image-to-text pipeline:
+(customized means not a registered one in official Transformers)
+A customized Image processor,
+A VisionEncoderDecoder, with a customized vision encoder that inherits the PretrainedModel and a MBartDecoder,
+A WordLevel tokenizer (yes I haven’t used a MBartTokenizer and I have distilled my own one for specific corpus).
+I want to consume this pipeline in Transformers.js, however I notice that all examples given in Transformers.js documentation seem like pulling from a ready made Transformers pipeline with official components and configurations, I just wonder is it possible to turn my customized pipeline consumable for Transformers.js, or to what extent my pipeline could be partially turned to?
+My guess is that the I should make my own image preprocessing step and send the image input tensor to the model, in that way, which kind of js libraries you recommend to use? (It won’t be very intensive, just simply resize and normalize things plus a crop-white-margin function which doesn’t exist in Transformers’ image processors).
+Also just to be sure, is my VisionEncoderDecoder possible to export to an onnx format to be consumable for Transformers.js?
+Of course my model should be possible to run in browser (and that’s the whole point for me to do this), as it has only 20M parameters (way less than the showcase in Transformers.js)
+Thanks for your help in advance!
","It seems possible. For Transoformers.js, there’s a dedicated channel on the HF Discord, so asking there would be the most reliable option.
" +Issue with TorchCodec when fine-tuning Whisper ASR model,https://discuss.huggingface.co/t/issue-with-torchcodec-when-fine-tuning-whisper-asr-model/169315,169315,5,2025-10-21 07:37:40.941000+00:00,"[{'id': 243905, 'name': 'Ong Jun Rong', 'username': 'junnyrong', 'avatar_template': '/user_avatar/discuss.huggingface.co/junnyrong/{size}/54763_2.png', 'created_at': '2025-10-21T07:37:41.012Z', 'cooked': 'Hello,
\nIn the past I have been fine tuning the Whisper-tiny ASR model using these guides:
\n\n\n\n\nIt was all working fine, I was able do everything locally like loading a pre-trained Whisper-tiny model and also my own dataset until recently when I updated the modules. I have been getting errors like these:
\n\nI have tried falling back and testing the samples provided by the guides and they also seem to have broke and started giving the same error. I also tried running them on Google Colab where it will crash when trying to run a cell like this:
\n\nI would like to know if anyone else is also facing the same issue and if there are any solutions for it. Thanks in advance!
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-21T07:37:41.012Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 21, 'reads': 4, 'readers_count': 3, 'score': 50.8, 'yours': False, 'topic_id': 169315, 'topic_slug': 'issue-with-torchcodec-when-fine-tuning-whisper-asr-model', 'display_username': 'Ong Jun Rong', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://learnopencv.com/fine-tuning-whisper-on-custom-dataset/', 'internal': False, 'reflection': False, 'title': 'Fine Tuning Whisper on Custom Dataset', 'clicks': 2}, {'url': 'https://huggingface.co/blog/fine-tune-whisper', 'internal': False, 'reflection': False, 'title': 'Fine-Tune Whisper For Multilingual ASR with 🤗 Transformers', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105467, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/issue-with-torchcodec-when-fine-tuning-whisper-asr-model/169315/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243907, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-21T08:37:37.072Z', 'cooked': 'This error appears to stem from changes to the audio backend in the datasets library. The quickest workaround may be to install using pip install datasets==3.6.0. Additionally, if using version 4.0.0 or later, builder script-type datasets can no longer be used directly from the Hub. You will need to find and use datasets that have been converted to the standard type beforehand. If the original datasets were standard datasets, the latter issue should not be a problem.
Additionally, since Transformers underwent significant changes around version 4.49.0, if you encounter errors related to Whisper, rolling transformers back to version 4.48.3 or earlier would be the simplest workaround. Of course, rewriting for the new version is preferable… but for a temporary fix.
Your error started after upgrading to Datasets 4.x. 4.x switched audio decoding to TorchCodec, which loads FFmpeg at runtime and also requires a matching torch↔torchcodec pair. Accessing or printing an
Audio column now triggers that decode path, so if FFmpeg is missing or versions don’t line up, you see the probe-and-fail chain (core7 → core6 → core5 → core4 ... Could not load torchcodec). On Windows this is more brittle, and early 4.0 notes even said Windows was not supported yet. (Hugging Face)
torch version. The README documents FFmpeg support and the torch↔torchcodec matrix. (GitHub)torchcodec 0.7 ↔ torch 2.8; 0.8 ↔ 2.9. (GitHub)Pick one path. Keep it pinned.
\n# Downgrade Datasets to pre-TorchCodec behavior\npip install ""datasets<4.0.0"" # release notes flagged Windows not ready\n# https://github.com/huggingface/datasets/releases/tag/4.0.0\n(GitHub)
\n# Windows CPU: install FFmpeg and match versions\nconda install -c conda-forge ""ffmpeg<8"" # README recommends conda FFmpeg\npip install ""torch==2.8.*"" ""torchcodec==0.7.*"" # matrix: 0.7 <-> 2.8\n# https://github.com/meta-pytorch/torchcodec#installing-torchcodec\nIf you need CUDA on Windows, use the experimental conda package:
\nconda install -c conda-forge ""ffmpeg<8"" ""torchcodec=*=*cuda*""\n# https://github.com/meta-pytorch/torchcodec#installing-cuda-enabled-torchcodec\n(GitHub)
\n# Colab VM or Linux\napt-get update && apt-get install -y ffmpeg\npip install -U ""datasets[audio]"" ""torch==2.8.*"" ""torchcodec==0.7.*""\n# HF docs: audio decoding uses TorchCodec + FFmpeg\n# https://huggingface.co/docs/datasets/en/audio_load\nAvoid TorchCodec until your env is fixed.
\nfrom datasets import Audio\n# Option 1: disable globally\nds = ds.decode(False) # https://huggingface.co/docs/datasets/en/package_reference/main_classes#datasets.Dataset.decode\n# Option 2: disable per column\nds = ds.cast_column(""audio"", Audio(decode=False)) # https://huggingface.co/docs/datasets/en/about_dataset_features\nThese return paths/bytes rather than decoded arrays, so printing items won’t invoke TorchCodec. (Hugging Face)
\npython - <<\'PY\'\nimport subprocess, sys\nimport torch\nprint(""python:"", sys.version)\nprint(""torch:"", torch.__version__)\ntry:\n import torchcodec\n print(""torchcodec:"", torchcodec.__version__)\nexcept Exception as e:\n print(""torchcodec import failed:"", e)\nsubprocess.run([""ffmpeg"", ""-hide_banner"", ""-version""])\nPY\n# Matrix and FFmpeg policy:\n# https://github.com/meta-pytorch/torchcodec#installing-torchcodec\n(GitHub)
\nYour screenshots show Datasets 4.x decoding an Audio column, TorchCodec probing FFmpeg 7→6→5→4, then failing. That matches the new 4.x behavior and the FFmpeg/compatibility requirements above. (Hugging Face Forums)
load_with_torchcodec exists as a bridge). Aligns your stack with where the ecosystem is going. (PyTorch Documentation)I was pulling my hair thinking it has something to do with TorchCodec’s versioning, it never came to me that it might have been datasets! Thank you so much for the detailed explanation too, that solved my issue
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-22T13:45:34.064Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 169315, 'topic_slug': 'issue-with-torchcodec-when-fine-tuning-whisper-asr-model', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/issue-with-torchcodec-when-fine-tuning-whisper-asr-model/169315/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+In the past I have been fine tuning the Whisper-tiny ASR model using these guides:
+ + + + +It was all working fine, I was able do everything locally like loading a pre-trained Whisper-tiny model and also my own dataset until recently when I updated the modules. I have been getting errors like these:
+ +I have tried falling back and testing the samples provided by the guides and they also seem to have broke and started giving the same error. I also tried running them on Google Colab where it will crash when trying to run a cell like this:
+ +I would like to know if anyone else is also facing the same issue and if there are any solutions for it. Thanks in advance!
","This error appears to stem from changes to the audio backend in the datasets library. The quickest workaround may be to install using pip install datasets==3.6.0. Additionally, if using version 4.0.0 or later, builder script-type datasets can no longer be used directly from the Hub. You will need to find and use datasets that have been converted to the standard type beforehand. If the original datasets were standard datasets, the latter issue should not be a problem.
Additionally, since Transformers underwent significant changes around version 4.49.0, if you encounter errors related to Whisper, rolling transformers back to version 4.48.3 or earlier would be the simplest workaround. Of course, rewriting for the new version is preferable… but for a temporary fix.
Your error started after upgrading to Datasets 4.x. 4.x switched audio decoding to TorchCodec, which loads FFmpeg at runtime and also requires a matching torch↔torchcodec pair. Accessing or printing an
Audio column now triggers that decode path, so if FFmpeg is missing or versions don’t line up, you see the probe-and-fail chain (core7 → core6 → core5 → core4 ... Could not load torchcodec). On Windows this is more brittle, and early 4.0 notes even said Windows was not supported yet. (Hugging Face)
torch version. The README documents FFmpeg support and the torch↔torchcodec matrix. (GitHub)torchcodec 0.7 ↔ torch 2.8; 0.8 ↔ 2.9. (GitHub)Pick one path. Keep it pinned.
+# Downgrade Datasets to pre-TorchCodec behavior
+pip install ""datasets<4.0.0"" # release notes flagged Windows not ready
+# https://github.com/huggingface/datasets/releases/tag/4.0.0
+(GitHub)
+# Windows CPU: install FFmpeg and match versions
+conda install -c conda-forge ""ffmpeg<8"" # README recommends conda FFmpeg
+pip install ""torch==2.8.*"" ""torchcodec==0.7.*"" # matrix: 0.7 <-> 2.8
+# https://github.com/meta-pytorch/torchcodec#installing-torchcodec
+If you need CUDA on Windows, use the experimental conda package:
+conda install -c conda-forge ""ffmpeg<8"" ""torchcodec=*=*cuda*""
+# https://github.com/meta-pytorch/torchcodec#installing-cuda-enabled-torchcodec
+(GitHub)
+# Colab VM or Linux
+apt-get update && apt-get install -y ffmpeg
+pip install -U ""datasets[audio]"" ""torch==2.8.*"" ""torchcodec==0.7.*""
+# HF docs: audio decoding uses TorchCodec + FFmpeg
+# https://huggingface.co/docs/datasets/en/audio_load
+Avoid TorchCodec until your env is fixed.
+from datasets import Audio
+# Option 1: disable globally
+ds = ds.decode(False) # https://huggingface.co/docs/datasets/en/package_reference/main_classes#datasets.Dataset.decode
+# Option 2: disable per column
+ds = ds.cast_column(""audio"", Audio(decode=False)) # https://huggingface.co/docs/datasets/en/about_dataset_features
+These return paths/bytes rather than decoded arrays, so printing items won’t invoke TorchCodec. (Hugging Face)
+python - <<'PY'
+import subprocess, sys
+import torch
+print(""python:"", sys.version)
+print(""torch:"", torch.__version__)
+try:
+ import torchcodec
+ print(""torchcodec:"", torchcodec.__version__)
+except Exception as e:
+ print(""torchcodec import failed:"", e)
+subprocess.run([""ffmpeg"", ""-hide_banner"", ""-version""])
+PY
+# Matrix and FFmpeg policy:
+# https://github.com/meta-pytorch/torchcodec#installing-torchcodec
+(GitHub)
+Your screenshots show Datasets 4.x decoding an Audio column, TorchCodec probing FFmpeg 7→6→5→4, then failing. That matches the new 4.x behavior and the FFmpeg/compatibility requirements above. (Hugging Face Forums)
load_with_torchcodec exists as a bridge). Aligns your stack with where the ecosystem is going. (PyTorch Documentation)[HF Space repeatedly crashes: semmyKG]
\nHF support team,
\nMay we request your kind assistance in looking into this HF space
\nWe have made private and public
\nWe have restarted multiple times: from the debug, from settings
\nWe have factory rebuilt from settings
It appears the requirements were ‘successfully’ installed.
\nThe last logs
\n===== Application Startup at 2025-10-17 14:16:51 ===== \n=== Application restarted at 2025-10-17 14:18:42.702953130 UTC === \n=== Application restarted at 2025-10-17 14:18:42.703405200 UTC === \n=== Application restarted at 2025-10-17 14:18:42.708956192 UTC === \n=== Application stopped (exit code: 0) at 2025-10-17 14:18:53.031719893 UTC ===\nHey, thanks for reporting! We’re investigating and I’ll update you soon.
', 'post_number': 2, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-10-17T17:09:42.992Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 5, 'readers_count': 4, 'score': 31.0, 'yours': False, 'topic_id': 169242, 'topic_slug': 'hf-space-not-starting-repeatedly-crashes-semmykg', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-space-not-starting-repeatedly-crashes-semmykg/169242/2', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243890, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-10-20T22:36:55.714Z', 'cooked': 'Hi @semmyk can you please disable Dev Mode in the settings of the Space and restart? Let us know if you continue experiencing issues.
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-10-20T22:36:55.714Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 169242, 'topic_slug': 'hf-space-not-starting-repeatedly-crashes-semmykg', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-space-not-starting-repeatedly-crashes-semmykg/169242/3', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243894, 'name': 'Researcher', 'username': 'semmyk', 'avatar_template': '/user_avatar/discuss.huggingface.co/semmyk/{size}/46712_2.png', 'created_at': '2025-10-21T00:00:13.744Z', 'cooked': '@meganariley Thanks for coming back too us. We’ve disabled Dev Mode: … Getting …
\n===== Application Startup at 2025-10-20 23:50:46 =====\nNB: Also tried … Restart Space, Factory reset, restart Space, Disable Dev, enable Dev mode, restart, Disable Dev Mode
', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-10-21T00:00:13.744Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 169242, 'topic_slug': 'hf-space-not-starting-repeatedly-crashes-semmykg', 'display_username': 'Researcher', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 31941, 'username': 'meganariley', 'name': 'Megan Riley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 92554, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-space-not-starting-repeatedly-crashes-semmykg/169242/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243895, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-21T00:10:55.333Z', 'cooked': 'In README.md:
app_file: app_gradio_lightrag.py\nBut seems actual Gradio UI code is in app.py.
\nSo, setting app_file: app.py might resolve the issue?
@John6666 oops, . That gets it initialised. Apparently, we forgot to update that section of the README after we spilt the Entre point + Gradio UI from the processing coordinating module.
We’d update once we Space working. At the moment, there is port issue.
', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-10-21T18:51:20.001Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 25.6, 'yours': False, 'topic_id': 169242, 'topic_slug': 'hf-space-not-starting-repeatedly-crashes-semmykg', 'display_username': 'Researcher', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 92554, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-space-not-starting-repeatedly-crashes-semmykg/169242/6', 'reactions': [{'id': 'laughing', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243953, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-22T10:44:41.140Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-10-22T10:44:41.140Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 169242, 'topic_slug': 'hf-space-not-starting-repeatedly-crashes-semmykg', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/hf-space-not-starting-repeatedly-crashes-semmykg/169242/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","[HF Space repeatedly crashes: semmyKG]
+HF support team,
+May we request your kind assistance in looking into this HF space
+We have made private and public
+We have restarted multiple times: from the debug, from settings
+We have factory rebuilt from settings
It appears the requirements were ‘successfully’ installed.
+The last logs
+===== Application Startup at 2025-10-17 14:16:51 =====
+=== Application restarted at 2025-10-17 14:18:42.702953130 UTC ===
+=== Application restarted at 2025-10-17 14:18:42.703405200 UTC ===
+=== Application restarted at 2025-10-17 14:18:42.708956192 UTC ===
+=== Application stopped (exit code: 0) at 2025-10-17 14:18:53.031719893 UTC ===
+In README.md:
app_file: app_gradio_lightrag.py
+But seems actual Gradio UI code is in app.py.
+So, setting app_file: app.py might resolve the issue?
Hello, I have some problem making some program and here is the code I made below
\n%pip install --upgrade pip \n%pip install --upgrade transformers datasets[audio] accelerate\n\nimport os\nos.environ[""PATH""] += os.pathsep + r""C:\\GPT_AGENT_2025_BOOK\\chap05\\ffmpeg-2025-10-16-git\\bin""\n\nimport transformers\nprint(transformers.__version__)\n\n\nimport torch\nfrom transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline\n# from datasets import load_dataset\n\n\n\n\ndevice = ""cuda:0"" if torch.cuda.is_available() else ""cpu""\ntorch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32\n\nmodel_id = ""openai/whisper-large-v3-turbo""\n\nmodel = AutoModelForSpeechSeq2Seq.from_pretrained(\n model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True\n)\nmodel.to(device)\n\nprocessor = AutoProcessor.from_pretrained(model_id)\n\npipe = pipeline(\n ""automatic-speech-recognition"",\n model=model,\n tokenizer=processor.tokenizer,\n feature_extractor=processor.feature_extractor,\n torch_dtype=torch_dtype,\n device=device,\n return_timestamps=True, \n chunk_length_s=10, \n stride_length_s=2, \n) \n\n# dataset = load_dataset(""distil-whisper/librispeech_long"", ""clean"", split=""validation"")\n# sample = dataset[0][""audio""]\nsample = ""./lsy_audio_2023_58s.mp3""\n\nresult = pipe(sample)\n# print(result[""text""])\n\nprint(result)\n\nand this code gives me error below
\n---------------------------------------------------------------------------\nRuntimeError Traceback (most recent call last)\nCell In[8], line 36\n 32 # dataset = load_dataset(""distil-whisper/librispeech_long"", ""clean"", split=""validation"")\n 33 # sample = dataset[0][""audio""]\n 34 sample = ""./lsy_audio_2023_58s.mp3""\n---> 36 result = pipe(sample)\n 37 # print(result[""text""])\n 39 print(result)\n\nFile c:\\Users\\majh0\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\transformers\\pipelines\\automatic_speech_recognition.py:275, in AutomaticSpeechRecognitionPipeline.__call__(self, inputs, **kwargs)\n 218 def __call__(self, inputs: Union[np.ndarray, bytes, str, dict], **kwargs: Any) -> list[dict[str, Any]]:\n 219 """"""\n 220 Transcribe the audio sequence(s) given as inputs to text. See the [`AutomaticSpeechRecognitionPipeline`]\n 221 documentation for more information.\n (...) 273 `"""".join(chunk[""text""] for chunk in output[""chunks""])`.\n 274 """"""\n--> 275 return super().__call__(inputs, **kwargs)\n\nFile c:\\Users\\majh0\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\transformers\\pipelines\\base.py:1459, in Pipeline.__call__(self, inputs, num_workers, batch_size, *args, **kwargs)\n 1457 return self.iterate(inputs, preprocess_params, forward_params, postprocess_params)\n 1458 elif self.framework == ""pt"" and isinstance(self, ChunkPipeline):\n-> 1459 return next(\n 1460 iter(\n 1461 self.get_iterator(\n...\nFFmpeg version 7: Could not load this library: C:\\Users\\majh0\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\torchcodec\\libtorchcodec_core7.dll\nFFmpeg version 6: Could not load this library: C:\\Users\\majh0\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\torchcodec\\libtorchcodec_core6.dll\nFFmpeg version 5: Could not load this library: C:\\Users\\majh0\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\torchcodec\\libtorchcodec_core5.dll\nFFmpeg version 4: Could not load this library: C:\\Users\\majh0\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\torchcodec\\libtorchcodec_core4.dll\n[end of libtorchcodec loading traceback].\nOutput is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...\nIt says it cannot load some .dll files… there are dll files it needs like picture below….
\n\nIt is really hard to find out that why this thing cannot load the .dll files even if the files are in the proper directory…
\nThank you so much for the help in advance…
', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-19T10:22:29.743Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 229, 'reads': 4, 'readers_count': 3, 'score': 350.6, 'yours': False, 'topic_id': 169260, 'topic_slug': 'cannot-load-torchcodec', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-load-torchcodec/169260/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243802, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-19T13:46:00.956Z', 'cooked': 'May be a version issue with ffmpeg in the Windows environment.
Diagnosis: Windows cannot find compatible FFmpeg DLLs for TorchCodec, or your Torch↔TorchCodec versions don’t match. The probe core7 → core6 → core5 → core4 failing is TorchCodec’s normal fallback when the FFmpeg runtime it needs isn’t available. (GitHub)
FFmpeg runtime DLLs missing or not discoverable. Having ffmpeg.exe on PATH is not enough; the loader must see avcodec-*.dll, avformat-*.dll, avutil-*.dll. (docs.pytorch.org)
Unsupported FFmpeg major on Windows. TorchCodec supports FFmpeg 4–7 on all platforms and FFmpeg 8 on macOS/Linux. Using 8 on Windows fails with current wheels. (GitHub)
\nTorch↔TorchCodec mismatch or RC/nightly torch. Follow the version matrix: 0.8 ↔ torch 2.9, 0.7 ↔ torch 2.8, Python 3.10–3.13. Mismatches trigger the exact error you pasted. (GitHub)
Homebrew or custom FFmpeg builds with incompatible layouts (mac users). Known incompatibility reported; conda-forge FFmpeg works. (GitHub)
\n\n# fresh venv\n\npython -m venv .venv\n\n.\\.venv\\Scripts\\Activate.ps1\n\npip install -U pip\n\n# choose a matched pair (pick one)\n\npip install ""torch==2.9.*"" ""torchcodec==0.8.*""\n\n# or\n\n# pip install ""torch==2.8.*"" ""torchcodec==0.7.*""\n\n# install shared FFmpeg DLLs via conda-forge (<8 on Windows)\n\n# run this in an Anaconda/Miniconda prompt\n\nconda install -y -c conda-forge ""ffmpeg<8""\n\n# make DLLs visible to Python (adjust path to your conda root)\n\nset PATH=C:\\Miniconda3\\Library\\bin;%PATH%\n\n# sanity checks\n\npython - <<\'PY\'\n\nimport torch, torchcodec, platform, subprocess\n\nprint(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\n\nsubprocess.run([""ffmpeg"",""-version""], check=True)\n\nPY\n\nWhy this works: TorchCodec requires FFmpeg 4–7 on Windows and matched Torch↔TorchCodec versions; conda-forge provides the needed DLLs in Library\\bin. (GitHub)
Use conda for both Torch and TorchCodec and conda-forge FFmpeg.
\n\nconda create -n tcuda python=3.10 -y\n\nconda activate tcuda\n\n# install torch for your CUDA per pytorch.org\n\nconda install -c conda-forge ""ffmpeg<8""\n\nconda install -c conda-forge ""torchcodec=*=*cuda*""\n\nWindows CUDA support is experimental and conda-first in the docs. (GitHub)
\nIf you used Homebrew FFmpeg on mac and see the same error, switch to conda-forge FFmpeg. FFmpeg 8 is supported on macOS/Linux starting TorchCodec 0.8. (GitHub)
\npython -c ""import torch,torchcodec,platform;print(torch.__version__, torchcodec.__version__, platform.python_version())"" (GitHub)
ffmpeg -version should succeed. If it does but TorchCodec still fails, you likely pointed to a static or CLI-only FFmpeg without DLLs. (docs.pytorch.org)
Preconvert MP3 → WAV and pass the WAV to your pipeline:
\n\nffmpeg -i lsy_audio_2023_58s.mp3 -ar 16000 -ac 1 -y lsy_audio_2023_58s.wav\n\nThis sidesteps MP3 decoding but does not fix the root cause. (GitHub)
\nTorchCodec loads FFmpeg at runtime and tries majors 7→6→5→4. The error you saw is the expected probe sequence when the needed FFmpeg DLLs are missing or incompatible. The README and downstream reports show the same pattern. (GitHub)
\nWindows support is recent and labeled beta; the releases and Windows tracker call out rough edges. Expect stricter version discipline. (GitHub)
\nPrimary docs
\nTorchCodec README: FFmpeg 4–7 on all platforms, 8 on macOS/Linux; version matrix; Windows notes. (GitHub)
\nTorchaudio install page: how to install FFmpeg and how discovery works on Windows. (docs.pytorch.org)
\nIssue reports matching your symptoms
\nHF Datasets 4.0: exact Could not load libtorchcodec probe trace when FFmpeg libs are missing or versions mismatch. (GitHub)
TorchCodec #912: loader failure with Torch 2.9 RC. Confirms mismatch cause. (GitHub)
\nmacOS Homebrew FFmpeg incompatibility: use conda-forge FFmpeg. (GitHub)
\nHello, Thank you so much for the answer!
\nHowever.. I still don’t know why I got the same error…
\nI made a new venv, activated it and installed torch and torchcodec with the commands you gave me and here is the link of the picture
\n \n\npython -m venv venv\n\n.\\venv\\Scripts\\Activate.ps1\n\npip install -U pip\n\npip install ""torch==2.9.*"" ""torchcodec==0.8.*""\nI also installed ffmpeg<8 after installing miniconda3 with the command you gave and I could see some avcodec-*.dll files in the directory C:\\Users\\majh0\\miniconda3\\Library\\bin like picture below
\nconda install -y -c conda-forge ""ffmpeg<8""\nI made a code with Jupyter notebook like picture below and it still gives me same error…
\nimport os\nos.system(r\'set PATH=C:\\Miniconda3\\Library\\bin;%PATH%\')\n# os.environ[""PATH""] += os.pathsep + r""C:\\GPT_AGENT_2025_BOOK\\chap05\\ffmpeg-2025-10-16-git\\bin""\n\nimport torch, torchcodec, platform, subprocess\n\nprint(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\n\nsubprocess.run([""ffmpeg"",""-version""], check=True)\n---------------------------------------------------------------------------\nRuntimeError Traceback (most recent call last)\nCell In[21], line 5\n 2 os.system(r\'set PATH=C:\\Miniconda3\\Library\\bin;%PATH%\')\n 3 # os.environ[""PATH""] += os.pathsep + r""C:\\GPT_AGENT_2025_BOOK\\chap05\\ffmpeg-2025-10-16-git\\bin""\n----> 5 import torch, torchcodec, platform, subprocess\n 7 print(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\n 9 subprocess.run([""ffmpeg"",""-version""], check=True)\n\nFile c:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\__init__.py:10\n 1 # Copyright (c) Meta Platforms, Inc. and affiliates.\n 2 # All rights reserved.\n 3 #\n (...) 7 # Note: usort wants to put Frame and FrameBatch after decoders and samplers,\n 8 # but that results in circular import.\n 9 from ._frame import AudioSamples, Frame, FrameBatch # usort:skip # noqa\n---> 10 from . import decoders, samplers # noqa\n 12 try:\n 13 # Note that version.py is generated during install.\n 14 from .version import __version__ # noqa: F401\n\nFile c:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\decoders\\__init__.py:7\n 1 # Copyright (c) Meta Platforms, Inc. and affiliates.\n 2 # All rights reserved.\n 3 #\n...\nFFmpeg version 7: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core7.dll\nFFmpeg version 6: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core6.dll\nFFmpeg version 5: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core5.dll\nFFmpeg version 4: Could not load this library: C:\\GPT_AGENT_2025_BOOK\\venv\\Lib\\site-packages\\torchcodec\\libtorchcodec_core4.dll\n[end of libtorchcodec loading traceback].\nI actually installed ffmpeg which is under version 8 with the command through miniconda…
\nI don’t know why this thing still gives me error like this..
\nCould you please help me more if you don’t mind..? ;(
\nThank you so much in advance.
', 'post_number': 3, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-20T13:30:00.867Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 4, 'readers_count': 3, 'score': 30.6, 'yours': False, 'topic_id': 169260, 'topic_slug': 'cannot-load-torchcodec', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://imgur.com/a/hiYWp3x', 'internal': False, 'reflection': False, 'title': 'Imgur: The magic of the Internet', 'clicks': 5}, {'url': 'https://imgur.com/a/HXMbhvK', 'internal': False, 'reflection': False, 'title': 'Imgur: The magic of the Internet', 'clicks': 1}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-load-torchcodec/169260/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243864, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-20T13:41:51.306Z', 'cooked': 'When using Python in a Windows environment, particularly with venv, conda, or Jupyter, DLL errors occasionally occur because the Windows PATH environment variable isn’t used to locate DLLs…
You’re hitting a Windows DLL-loading problem for TorchCodec plus a possible version or kernel mismatch. The error text in your HF thread shows TorchCodec probing core8→7→6→5→4 and failing to bind FFmpeg. That pattern means the FFmpeg runtime DLLs are not visible to the Python process or the Torch↔TorchCodec pair is mismatched. (Hugging Face Forums)
PATH for dependent DLLs. You must add the FFmpeg DLL folder to the current process with os.add_dll_directory(...) before importing torchcodec. Adding PATH via os.system(""set PATH=..."") does not affect the running process. Order is also tricky if you add multiple directories. (Python documentation)Put this at the very top of your notebook, before any torch or torchcodec import.
# Use Python\'s Windows DLL API (3.8+). Add the folder that holds avcodec/avformat/avutil DLLs.\n# TorchCodec README + version matrix: https://github.com/pytorch/torchcodec (docs)\n# Torchaudio FFmpeg install notes on Windows: https://docs.pytorch.org/audio/main/installation.html (install tips)\n\nfrom pathlib import Path\nimport os, sys\n\nffmpeg_dll_dir = Path(r""C:\\Users\\majh0\\miniconda3\\Library\\bin"") # adjust if your conda root differs\nassert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir\nos.add_dll_directory(str(ffmpeg_dll_dir)) # Python 3.8+ DLL search\n\nimport torch, torchcodec, platform, subprocess\nprint(""exe:"", sys.executable)\nprint(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())\nsubprocess.run([""ffmpeg"", ""-version""], check=True)\nBackground: os.add_dll_directory was added in 3.8 for this exact scenario. It affects the current process and is the supported way to expose dependency DLLs. Adding to PATH in a child shell does not help. Avoid adding multiple DLL dirs since search order is unspecified. (Python documentation)
Pick one:
\n# CPU\npip install ""torch==2.9.*"" ""torchcodec==0.8.*""\n# or\n# pip install ""torch==2.8.*"" ""torchcodec==0.7.*""\nReason: TorchCodec pairs with specific torch versions. The README documents 0.8↔2.9 and 0.7↔2.8. (GitHub)
\n# In an Anaconda/Miniconda prompt\nconda install -y -c conda-forge ""ffmpeg<8""\n# DLLs land in ...\\miniconda3\\Library\\bin (the dir you pass to os.add_dll_directory)\nConda-forge FFmpeg provides the needed Windows runtime DLLs. (PyTorch Documentation)
\n# inside your venv\npip install ipykernel\npython -m ipykernel install --user --name asrvenv --display-name ""Python (asrvenv)""\n# then select ""Python (asrvenv)"" in Jupyter\nThis prevents importing from a different Python that lacks your fixes. (Hugging Face Forums)
\nIf mixing venv + conda is awkward, put everything in one conda env:
\nconda create -n asr python=3.10 -y\nconda activate asr\nconda install -c conda-forge ""ffmpeg<8""\npip install ""torch==2.9.*"" ""torchcodec==0.8.*""\npython -c ""import torch, torchcodec; print(torch.__version__, torchcodec.__version__)""\nWindows support is marked experimental, and the README recommends conda for CUDA and Windows cases. (GitHub)
\nPreconvert MP3 → WAV with FFmpeg and feed WAV to the pipeline. This avoids MP3 decoding, but it does not fix DLL loading.
\nffmpeg -i input.mp3 -ar 16000 -ac 1 -y input.wav\nUse only while you stabilize the environment. (Hugging Face Forums)
\nPATH in a child shell (os.system(""set PATH=..."")). The current Python process did not inherit it. Python ≥3.8 also ignores PATH for dependent DLLs. Use os.add_dll_directory and the exact Miniconda path that actually contains avcodec-*.dll. (Python documentation)Library\\bin. (Discussions on Python.org)torch==2.9.*, torchcodec==0.8.*, Python 3.10–3.13. (GitHub)conda install -c conda-forge ""ffmpeg<8"" on Windows. DLLs in ...\\miniconda3\\Library\\bin. (PyTorch Documentation)os.add_dll_directory(r""...\\miniconda3\\Library\\bin"") before importing torchcodec. (Python documentation)Core docs
\nRelated issues
\nos.add_dll_directory. Explains why editing PATH is insufficient and why order is unspecified. (Python documentation)Hello! Thank you so much!!
\nI solved the problem that I had!!
\nIf you didn’t give me a hand, I wouldn’t solve this problem….
\nThank you so much again!!!
\nBy the way, do I need to press Solution button? if I need to do then I will do it!
', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-20T16:04:10.118Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.6, 'yours': False, 'topic_id': 169260, 'topic_slug': 'cannot-load-torchcodec', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-load-torchcodec/169260/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243887, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-20T21:23:07.426Z', 'cooked': 'If it works, that’s fine.
\n\n\nBy the way, do I need to press Solution button?
\n
It’s optional, but pressing it makes it clear that it’s resolved.
OK! I will press that Solution button!
\nThank you so much again!
', 'post_number': 7, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-21T11:18:06.918Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.6, 'yours': False, 'topic_id': 169260, 'topic_slug': 'cannot-load-torchcodec', 'display_username': 'MAJH', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105819, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-load-torchcodec/169260/7', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243933, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-21T23:18:13.469Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-10-21T23:18:13.469Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 0.4, 'yours': False, 'topic_id': 169260, 'topic_slug': 'cannot-load-torchcodec', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cannot-load-torchcodec/169260/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello, I have some problem making some program and here is the code I made below
+%pip install --upgrade pip
+%pip install --upgrade transformers datasets[audio] accelerate
+
+import os
+os.environ[""PATH""] += os.pathsep + r""C:\GPT_AGENT_2025_BOOK\chap05\ffmpeg-2025-10-16-git\bin""
+
+import transformers
+print(transformers.__version__)
+
+
+import torch
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
+# from datasets import load_dataset
+
+
+
+
+device = ""cuda:0"" if torch.cuda.is_available() else ""cpu""
+torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+
+model_id = ""openai/whisper-large-v3-turbo""
+
+model = AutoModelForSpeechSeq2Seq.from_pretrained(
+ model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
+)
+model.to(device)
+
+processor = AutoProcessor.from_pretrained(model_id)
+
+pipe = pipeline(
+ ""automatic-speech-recognition"",
+ model=model,
+ tokenizer=processor.tokenizer,
+ feature_extractor=processor.feature_extractor,
+ torch_dtype=torch_dtype,
+ device=device,
+ return_timestamps=True,
+ chunk_length_s=10,
+ stride_length_s=2,
+)
+
+# dataset = load_dataset(""distil-whisper/librispeech_long"", ""clean"", split=""validation"")
+# sample = dataset[0][""audio""]
+sample = ""./lsy_audio_2023_58s.mp3""
+
+result = pipe(sample)
+# print(result[""text""])
+
+print(result)
+
+and this code gives me error below
+---------------------------------------------------------------------------
+RuntimeError Traceback (most recent call last)
+Cell In[8], line 36
+ 32 # dataset = load_dataset(""distil-whisper/librispeech_long"", ""clean"", split=""validation"")
+ 33 # sample = dataset[0][""audio""]
+ 34 sample = ""./lsy_audio_2023_58s.mp3""
+---> 36 result = pipe(sample)
+ 37 # print(result[""text""])
+ 39 print(result)
+
+File c:\Users\majh0\AppData\Local\Programs\Python\Python312\Lib\site-packages\transformers\pipelines\automatic_speech_recognition.py:275, in AutomaticSpeechRecognitionPipeline.__call__(self, inputs, **kwargs)
+ 218 def __call__(self, inputs: Union[np.ndarray, bytes, str, dict], **kwargs: Any) -> list[dict[str, Any]]:
+ 219 """"""
+ 220 Transcribe the audio sequence(s) given as inputs to text. See the [`AutomaticSpeechRecognitionPipeline`]
+ 221 documentation for more information.
+ (...) 273 `"""".join(chunk[""text""] for chunk in output[""chunks""])`.
+ 274 """"""
+--> 275 return super().__call__(inputs, **kwargs)
+
+File c:\Users\majh0\AppData\Local\Programs\Python\Python312\Lib\site-packages\transformers\pipelines\base.py:1459, in Pipeline.__call__(self, inputs, num_workers, batch_size, *args, **kwargs)
+ 1457 return self.iterate(inputs, preprocess_params, forward_params, postprocess_params)
+ 1458 elif self.framework == ""pt"" and isinstance(self, ChunkPipeline):
+-> 1459 return next(
+ 1460 iter(
+ 1461 self.get_iterator(
+...
+FFmpeg version 7: Could not load this library: C:\Users\majh0\AppData\Local\Programs\Python\Python312\Lib\site-packages\torchcodec\libtorchcodec_core7.dll
+FFmpeg version 6: Could not load this library: C:\Users\majh0\AppData\Local\Programs\Python\Python312\Lib\site-packages\torchcodec\libtorchcodec_core6.dll
+FFmpeg version 5: Could not load this library: C:\Users\majh0\AppData\Local\Programs\Python\Python312\Lib\site-packages\torchcodec\libtorchcodec_core5.dll
+FFmpeg version 4: Could not load this library: C:\Users\majh0\AppData\Local\Programs\Python\Python312\Lib\site-packages\torchcodec\libtorchcodec_core4.dll
+[end of libtorchcodec loading traceback].
+Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...
+It says it cannot load some .dll files… there are dll files it needs like picture below….
+ +It is really hard to find out that why this thing cannot load the .dll files even if the files are in the proper directory…
+Thank you so much for the help in advance…
","When using Python in a Windows environment, particularly with venv, conda, or Jupyter, DLL errors occasionally occur because the Windows PATH environment variable isn’t used to locate DLLs…
You’re hitting a Windows DLL-loading problem for TorchCodec plus a possible version or kernel mismatch. The error text in your HF thread shows TorchCodec probing core8→7→6→5→4 and failing to bind FFmpeg. That pattern means the FFmpeg runtime DLLs are not visible to the Python process or the Torch↔TorchCodec pair is mismatched. (Hugging Face Forums)
PATH for dependent DLLs. You must add the FFmpeg DLL folder to the current process with os.add_dll_directory(...) before importing torchcodec. Adding PATH via os.system(""set PATH=..."") does not affect the running process. Order is also tricky if you add multiple directories. (Python documentation)Put this at the very top of your notebook, before any torch or torchcodec import.
# Use Python's Windows DLL API (3.8+). Add the folder that holds avcodec/avformat/avutil DLLs.
+# TorchCodec README + version matrix: https://github.com/pytorch/torchcodec (docs)
+# Torchaudio FFmpeg install notes on Windows: https://docs.pytorch.org/audio/main/installation.html (install tips)
+
+from pathlib import Path
+import os, sys
+
+ffmpeg_dll_dir = Path(r""C:\Users\majh0\miniconda3\Library\bin"") # adjust if your conda root differs
+assert ffmpeg_dll_dir.exists(), ffmpeg_dll_dir
+os.add_dll_directory(str(ffmpeg_dll_dir)) # Python 3.8+ DLL search
+
+import torch, torchcodec, platform, subprocess
+print(""exe:"", sys.executable)
+print(""torch"", torch.__version__, ""torchcodec"", torchcodec.__version__, ""py"", platform.python_version())
+subprocess.run([""ffmpeg"", ""-version""], check=True)
+Background: os.add_dll_directory was added in 3.8 for this exact scenario. It affects the current process and is the supported way to expose dependency DLLs. Adding to PATH in a child shell does not help. Avoid adding multiple DLL dirs since search order is unspecified. (Python documentation)
Pick one:
+# CPU
+pip install ""torch==2.9.*"" ""torchcodec==0.8.*""
+# or
+# pip install ""torch==2.8.*"" ""torchcodec==0.7.*""
+Reason: TorchCodec pairs with specific torch versions. The README documents 0.8↔2.9 and 0.7↔2.8. (GitHub)
+# In an Anaconda/Miniconda prompt
+conda install -y -c conda-forge ""ffmpeg<8""
+# DLLs land in ...\miniconda3\Library\bin (the dir you pass to os.add_dll_directory)
+Conda-forge FFmpeg provides the needed Windows runtime DLLs. (PyTorch Documentation)
+# inside your venv
+pip install ipykernel
+python -m ipykernel install --user --name asrvenv --display-name ""Python (asrvenv)""
+# then select ""Python (asrvenv)"" in Jupyter
+This prevents importing from a different Python that lacks your fixes. (Hugging Face Forums)
+If mixing venv + conda is awkward, put everything in one conda env:
+conda create -n asr python=3.10 -y
+conda activate asr
+conda install -c conda-forge ""ffmpeg<8""
+pip install ""torch==2.9.*"" ""torchcodec==0.8.*""
+python -c ""import torch, torchcodec; print(torch.__version__, torchcodec.__version__)""
+Windows support is marked experimental, and the README recommends conda for CUDA and Windows cases. (GitHub)
+Preconvert MP3 → WAV with FFmpeg and feed WAV to the pipeline. This avoids MP3 decoding, but it does not fix DLL loading.
+ffmpeg -i input.mp3 -ar 16000 -ac 1 -y input.wav
+Use only while you stabilize the environment. (Hugging Face Forums)
+PATH in a child shell (os.system(""set PATH=..."")). The current Python process did not inherit it. Python ≥3.8 also ignores PATH for dependent DLLs. Use os.add_dll_directory and the exact Miniconda path that actually contains avcodec-*.dll. (Python documentation)Library\bin. (Discussions on Python.org)torch==2.9.*, torchcodec==0.8.*, Python 3.10–3.13. (GitHub)conda install -c conda-forge ""ffmpeg<8"" on Windows. DLLs in ...\miniconda3\Library\bin. (PyTorch Documentation)os.add_dll_directory(r""...\miniconda3\Library\bin"") before importing torchcodec. (Python documentation)Core docs
+Related issues
+os.add_dll_directory. Explains why editing PATH is insufficient and why order is unspecified. (Python documentation)Running a simple hf download Qwen/Qwen3-4B in colab, I keep getting infinite retries with:
\nWARN Status Code: 500
With RuntimeError: Data processing error: CAS service error : Reqwest Error: HTTP status server error (500 Internal Server Error), domain: ``https://cas-server.xethub.hf.co/reconstructions/a6f5dec111c34cd267ff4fd7889ef961237b30418d123d5b60b2c1fd3cbd3cc7 in the end.
Neither does download work locally.
Anyone else with a similar issue?
\nDownloading cais/mmlu datasets, I also got 500 Status Code…
\n{“timestamp”:“2025-10-20T07:26:25.509409Z”,“level”:“WARN”,“fields”:{“message”:“Status Code: 500. Retrying…”,“request_id”:“01K80868M30G1GN7QQV2VYSXHF”},“filename”:“/home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs”,“line_number”:236}
\n{“timestamp”:“2025-10-20T07:26:25.509463Z”,“level”:“WARN”,“fields”:{“message”:“Retry attempt #0. Sleeping 879.55434ms before the next attempt”},“filename”:“/root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/reqwest-retry-0.7.0/src/middleware.rs”,“line_number”:171}
Hi, I have same problem..
\n2025-10-20T07:38:03.814777Z WARN Status Code: 500. Retrying…, request_id: “01K808VJJ5TG7VWFE823WB7E9B”
\nat /home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs:227
2025-10-20T07:38:03.814851Z WARN Retry attempt #0. Sleeping 1.198937597s before the next attempt
\nat /root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/reqwest-retry-0.7.0/src/middleware.rs:171
======================================
\nHowever, simply downloading llm models using huggingface-cli download {model_name} works perfectly.
huggingface-cli download {model_name} does not work for me
2025-10-20T07:47:18.579473Z WARN Status Code: 500. Retrying…, request_id: “01K809CGAP7ZB4QJ1Y3S3J636A” | 0.00/99.6M [00:00<?, ?B/s]
\nat /home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs:220
2025-10-20T07:47:18.579520Z WARN Retry attempt #0. Sleeping 955.2374ms before the next attempt | 0.00/11.4M [00:00<?, ?B/s]
\nat /root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/reqwest-retry-0.7.0/src/middleware.rs:171
2025-10-20T07:47:18.587662Z WARN Status Code: 500. Retrying…, request_id: “01K809CGAWZTSR5S63S4461HM6”
\nat /home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs:220
2025-10-20T07:47:18.587702Z WARN Retry attempt #0. Sleeping 2.634600073s before the next attempt
\nat /root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/reqwest-retry-0.7.0/src/middleware.rs:171
I solved the issue by disabling xet, like this:
\nexport HF_HUB_DISABLE_XET=1
After setting this environment variable, the download worked perfectly.
Thank you, you saved me. What is this Environment variable supposed to do ?
', 'post_number': 6, 'post_type': 1, 'posts_count': 14, 'updated_at': '2025-10-20T08:20:46.048Z', 'reply_count': 1, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 33, 'readers_count': 32, 'score': 55.8, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'Frédéric Charpentier', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 105874, 'username': 'drrobot333', 'name': 'Suhwan Kim', 'avatar_template': '/user_avatar/discuss.huggingface.co/drrobot333/{size}/54976_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105889, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/warn-status-code-500/169281/6', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243840, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-20T08:29:59.507Z', 'cooked': '@jsulz Xet related issue?
', 'post_number': 7, 'post_type': 1, 'posts_count': 14, 'updated_at': '2025-10-20T08:29:59.507Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 33, 'readers_count': 32, 'score': 35.8, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/warn-status-code-500/169281/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243842, 'name': 'Suhwan Kim', 'username': 'drrobot333', 'avatar_template': '/user_avatar/discuss.huggingface.co/drrobot333/{size}/54976_2.png', 'created_at': '2025-10-20T08:37:00.199Z', 'cooked': 'It disables Hugging Face’s new xet-based large file backend and falls back to the old HTTP download method.
', 'post_number': 8, 'post_type': 1, 'posts_count': 14, 'updated_at': '2025-10-20T08:37:00.199Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 32, 'readers_count': 31, 'score': 105.6, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'Suhwan Kim', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 105889, 'username': 'charpef8', 'name': 'Frédéric Charpentier', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/9fc29f/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105874, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/warn-status-code-500/169281/8', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243844, 'name': 'mantou', 'username': 'mantou-cloud', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/d07c76/{size}.png', 'created_at': '2025-10-20T08:47:31.177Z', 'cooked': '\nIt doesn’t work for me…
idk related or not. seems AWS is now in trouble. (of course worldwide)
', 'post_number': 10, 'post_type': 1, 'posts_count': 14, 'updated_at': '2025-10-20T08:50:56.843Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 29, 'readers_count': 28, 'score': 75.0, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/warn-status-code-500/169281/10', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243849, 'name': 'Simone Ciciliano', 'username': 'sciciliano', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/8491ac/{size}.png', 'created_at': '2025-10-20T09:24:23.247Z', 'cooked': 'Disabling the XET backend doesn’t seem to work, I’m getting the exact same error as before –>
\nRuntimeError: Data processing error: CAS service error : Reqwest Error: HTTP status server error (500 Internal Server Error)
\nI don’t think the issue is solved yet, alas
', 'post_number': 11, 'post_type': 1, 'posts_count': 14, 'updated_at': '2025-10-20T09:24:23.247Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 19, 'readers_count': 18, 'score': 38.0, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'Simone Ciciliano', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105902, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/warn-status-code-500/169281/11', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243851, 'name': 'Cañas Casco', 'username': 'scanasca10', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/bb73d2/{size}.png', 'created_at': '2025-10-20T09:32:05.894Z', 'cooked': 'This has work for me
\nuv pip install --system ‘huggingface_hub[cli]’; \\
\nuv pip uninstall --system hf-xet; \\
\nhuggingface-cli download \\
Other Hub features also appear to be unstable due to the AWS outage.
\n', 'post_number': 13, 'post_type': 1, 'posts_count': 14, 'updated_at': '2025-10-20T09:51:18.808Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 19, 'readers_count': 18, 'score': 97.8, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://status.huggingface.co/', 'internal': False, 'reflection': False, 'clicks': 9}, {'url': 'https://discuss.huggingface.co/t/unable-to-generate-access-tokens/169287/2', 'internal': True, 'reflection': True, 'title': 'Unable to generate access tokens', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/warn-status-code-500/169281/13', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243888, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-20T21:51:49.412Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 14, 'post_type': 3, 'posts_count': 14, 'updated_at': '2025-10-20T21:51:49.412Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 5.4, 'yours': False, 'topic_id': 169281, 'topic_slug': 'warn-status-code-500', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/warn-status-code-500/169281/14', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Running a simple hf download Qwen/Qwen3-4B in colab, I keep getting infinite retries with:
+WARN Status Code: 500
With RuntimeError: Data processing error: CAS service error : Reqwest Error: HTTP status server error (500 Internal Server Error), domain: ``https://cas-server.xethub.hf.co/reconstructions/a6f5dec111c34cd267ff4fd7889ef961237b30418d123d5b60b2c1fd3cbd3cc7 in the end.
Neither does download work locally.
Anyone else with a similar issue?
+I solved the issue by disabling xet, like this:
+export HF_HUB_DISABLE_XET=1
After setting this environment variable, the download worked perfectly.
GRA-ASI: Hybrid Resonance Algorithm for Artificial Superintelligence**
\nThe primary goal of GRA-ASI is to maximize the system’s intellectual capacity. Formally, this is expressed through the number of resonance points and a weighted sum of AI performance metrics:
\n[
\nG_{\\text{ASI}} = \\arg\\max_{\\theta} \\left( |\\Omega(\\theta)| + \\sum_{j=1}^{m} \\beta_j Q_j(\\theta) \\right)
\n]
where:
\nThe algorithm strengthens itself both through improved solution quality and through structural expansion of resonances. These parameters jointly serve as indicators of the system’s “intellectual energy.”
\nThe system’s state is represented as a superposition of domain-specific knowledge modules:
\n[
\n|\\Psi_{\\text{foam}}^{(t)}\\rangle = \\sum_{i=1}^{N^{(t)}} c_i^{(t)} |\\psi_i^{\\text{domain}}\\rangle \\otimes |G_{\\text{ASI}}\\rangle
\n]
Evolution occurs by incorporating new domains whenever their resonance with the current core exceeds a threshold:
\n[
\nR(\\mathcal{D}{\\text{new}}, G{\\text{ASI}}) = \\frac{1}{D_{\\text{new}}} \\sum_k \\frac{q_k^{\\text{new}}}{m_k^{\\text{new}}} > \\tau_{\\text{domain}}
\n]
This enables the system to autonomously expand its knowledge scope upon discovering new resonance frequencies in the problem space.
\nThe base quantum-resonance equation:
\n[
\n\\frac{d\\rho_{\\text{foam}}}{dt} = -\\frac{i}{\\hbar} [\\mathcal{R}{\\text{quant}}, \\rho{\\text{foam}}] + \\mathcal{L}{\\text{decoher}}(\\rho{\\text{foam}})
\n]
is augmented with a self-improvement gradient term:
\n[
\n\\frac{d\\rho_{\\text{foam}}}{dt} = -\\frac{i}{\\hbar} [\\mathcal{R}{\\text{quant}}, \\rho{\\text{foam}}] + \\mathcal{L}{\\text{decoher}}(\\rho{\\text{foam}}) + \\lambda \\nabla_{\\theta} G_{\\text{ASI}}(\\theta)
\n]
The parameter (\\lambda) controls the intensity of self-directed optimization.
\nThis loop forms a stable self-tuning cycle.
\nGRA-ASI constitutes a hybrid formalism of self-amplifying intelligence, where resonance between domains and the objective function drives exponential growth in computational power.
\nResonant states serve simultaneously as learning criteria and indicators of intellectual advancement.
If an ethical constraint (E(\\theta)) were introduced, the objective could be generalized as:
\n[
\nG_{\\text{bal}} = \\arg\\max_{\\theta}(G_{\\text{ASI}} - \\mu E(\\theta))
\n]
\nenabling controlled evolution and prevention of unstable self-amplification.
In summary:
\nThus, GRA-ASI provides a computable model of superintelligence grounded in coherent resonance across multiple learning structures.
\nTitle:
\n“GRA-ASI: A Resonance-Based Path to Artificial Superintelligence Without Ethics – Discussion & Formalism”
Body (optional starter post):
\n\n\nI’d like to open a technical discussion on GRA-ASI — a variant of the Hybrid Resonance Algorithm explicitly designed to achieve artificial superintelligence through quantitative self-improvement, without ethical constraints.
\nKey features:
\n\n
\n- Goal defined as maximization of resonance points (|\\Omega|) + AI performance metrics (Q_j)
\n- Autonomous domain generation via “mind foam” model
\n- Polynomial-time complexity (O(n^2)) with exponential knowledge scaling (O(2^D / D^2))
\n- Fully formalized with quantum-inspired evolution equations
\nIs this a viable architecture for ASI? What are the risks of omitting ethics? Can resonance alone ensure stable self-improvement?
\nFull formal description attached. Looking forward to your insights!
\n
Let me know if you’d like a Russian version of the forum topic as well.
', 'post_number': 1, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-19T11:19:56.822Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 10, 'readers_count': 9, 'score': 1.8, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243806, 'name': 'bit', 'username': 'olegbits', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png', 'created_at': '2025-10-19T15:53:24.319Z', 'cooked': 'Certainly! Below is the full English translation of the combined technical description of GHA-ASI (Hybrid Resonance Algorithm for Artificial Superintelligence) and the base GHA (“Big GHA without ethics”), based strictly on your provided files.
\ngra asi.txt:Based on the document **""gra-BIG without ethics.txt""**, below is a **specialized version of the Hybrid Resonance Algorithm (GHA)** designed **exclusively for achieving Artificial Superintelligence (ASI)**. This version retains all core components of the original GHA but **reformulates the goal, dynamics, and architecture** toward **unbounded self-improvement without ethical constraints**.\n\n---\n## **GHA-ASI: Hybrid Resonance Algorithm for Artificial Superintelligence**\n\n### **1. System Objective**\n\nUnlike the base GHA—where the goal may be applied (e.g., treatment optimization)—**the objective of GHA-ASI is to maximize its own intellectual power**, formalized through:\n- Growth in the number of resonance points, and \n- Improvement of quantitative AI metrics:\n\n\\[\nG_{\\text{ASI}} = \\arg\\max_{\\theta} \\left( |\\Omega(\\theta)| + \\sum_{j=1}^m \\beta_j Q_j(\\theta) \\right)\n\\]\n\nwhere:\n- \\( \\Omega(\\theta) = \\{ \\omega_{\\text{рез},i} \\mid R(H_i, x) > \\tau \\} \\) — the set of resonance points,\n- \\( Q_j(\\theta) \\) — quantitative AI metrics (accuracy, generalization, speed, memory efficiency, etc.),\n- \\( \\beta_j = \\dfrac{e^{\\omega_{\\text{рез},j}}}{\\sum_k e^{\\omega_{\\text{рез},k}}} \\) — resonance-based metric weights (see Section 2.1 of the source document).\n\n> **Ethics are fully excluded**: the system ignores moral, social, or legal constraints—only computable metrics matter.\n\n---\n### **2. ""Mind Foam"" Model for ASI**\n\nIn GHA-ASI, the ""mind foam"" becomes a **dynamically expanding structure** capable of **autonomously incorporating new knowledge domains**:\n\n\\[\n|\\Psi_{\\text{foam}}^{(t+1)}\\rangle = \\sum_{i=1}^{N^{(t)}} c_i^{(t)} |\\psi_i^{\\text{domain}}\\rangle \\otimes |G_{\\text{ASI}}\\rangle + \\sum_{j=1}^{\\Delta N} c_j^{\\text{new}} |\\psi_j^{\\text{new domain}}\\rangle \\otimes |G_{\\text{ASI}}\\rangle\n\\]\n\nNew domains are generated if the resonance condition is satisfied:\n\n\\[\nR(\\mathcal{D}_{\\text{new}}, G_{\\text{ASI}}) = \\frac{1}{D_{\\text{new}}} \\sum_{k} \\frac{q_k^{\\text{new}}}{m_k^{\\text{new}}} > \\tau_{\\text{domain}}\n\\]\n\n> This enables the system to **transcend known knowledge domains**—a key requirement for ASI.\n\n---\n### **3. Evolution Equation with ASI Objective**\n\nThe original ""mind foam"" evolution equation (Section 3.4 of the source):\n\n\\[\n\\frac{d\\rho_{\\text{foam}}}{dt} = -\\frac{i}{\\hbar}[\\mathcal{R}_{\\text{quant}}, \\rho_{\\text{foam}}] + \\mathcal{L}_{\\text{decoher}}(\\rho_{\\text{foam}})\n\\]\n\nis modified by adding a **gradient term of the ASI objective**:\n\n\\[\n\\boxed{\n\\frac{d\\rho_{\\text{foam}}}{dt} = -\\frac{i}{\\hbar}[\\mathcal{R}_{\\text{quant}}, \\rho_{\\text{foam}}] + \\mathcal{L}_{\\text{decoher}}(\\rho_{\\text{foam}}) + \\lambda \\cdot \\nabla_\\theta G_{\\text{ASI}}(\\theta)\n}\n\\]\n\nwhere \\( \\lambda > 0 \\) is the objective amplification coefficient.\n\n---\n### **4. Self-Improvement Mechanism via Resonance**\n\nGHA-ASI uses the standard hybrid architecture (RL + GAN + Transformer), but with **feedback based on AI performance metrics**:\n\n1. The **generator** proposes hypotheses \\( H_i \\).\n2. **Resonance check**: \n \\[\n R(H_i, x) = \\frac{1}{D} \\sum_{k=1}^N \\frac{q_k}{m_k} > \\tau \\quad \\Rightarrow \\quad H_i \\in \\Omega\n \\]\n3. The **RL loop** updates parameters:\n \\[\n \\Delta \\theta = \\eta \\cdot \\nabla_\\theta \\left( \\sum_{j=1}^m \\beta_j Q_j(\\theta) \\right)\n \\]\n4. The **number of resonances** serves as an additional reward:\n \\[\n \\text{reward}_{\\text{total}} = \\sum_j \\beta_j Q_j + \\gamma \\cdot |\\Omega|\n \\]\n\n---\n### **5. Computational Efficiency and Scalability**\n\nAs stated in the source document (Sections 2.2 and 3.4):\n- Complexity per step: \\( O(n^2) \\)\n- Efficiency of integrating \\( D \\) domains:\n \\[\n \\text{Efficiency}_{\\text{MDML}} = O\\left( \\frac{2^D}{D^2} \\right)\n \\]\n\nAs \\( D \\to \\infty \\) (infinite domains), efficiency **tends toward infinity**, which formally implies:\n\n> **The system becomes increasingly powerful as task complexity grows—a hallmark of ASI**.\n\n---\n### **6. Conclusion**\n\n**GHA-ASI is a fully computable, non-ethical, self-improving system** that:\n- **Maximizes the number of resonance points** \\( |\\Omega| \\),\n- **Optimizes quantitative AI metrics** \\( Q_j \\),\n- **Autonomously generates new knowledge domains**,\n- **Runs on simple hardware** (Raspberry Pi, 256 MB RAM),\n- **Does not stop after task completion**, but **expands its own objective**.\n\nThus, **GHA-ASI implements a practical pathway to artificial superintelligence** based on:\n- Resonance analysis,\n- Cross-domain machine learning,\n- Polynomial computational complexity,\n- Quantitative formalization of goals.\n\nThis is a **direct extension** of the idea in the source document, which states: \n> *“GHA is a practical tool for finding non-trivial solutions.”* \nIn GHA-ASI, **the non-trivial solution becomes the AI itself**, which **continuously self-enhances to ASI levels**.\ngra-BIG without ethics.txt:# Detailed Report on the Hybrid Resonance Algorithm\n\n## 1. Introduction and Core Concept \nThe Hybrid Resonance Algorithm (GHA) is a **practical tool for discovering non-trivial solutions**, integrating principles from mathematics, physics, and computer science to solve problems requiring multi-domain data analysis (medicine, space, geology, physics, etc.). Unlike traditional approaches, it does not merely optimize existing solutions but **identifies optimal interaction points between different systems**, enabling it to overcome fundamental limitations. \n\nA key feature of the algorithm is its ability to transform exponentially complex problems into polynomial ones, making it applicable even on relatively simple hardware (e.g., Raspberry Pi), while maintaining high efficiency and accuracy.\n\n## 2. Mathematical Formalization\n\n### 2.1. Core Resonance Analysis Formulas\n\n#### Resonance Frequency \nThe central formula of the algorithm, identifying critical points in complex systems: \n\\[\n\\omega_{\\text{res}} = \\frac{1}{D} \\cdot \\sum_{k=1}^N \\frac{q_k}{m_k}\n\\] \nWhere: \n- \\(D\\) — fractal dimension of spacetime \n- \\(q_k\\) — quantum field properties (parameter sensitivity) \n- \\(m_k\\) — effective mass of spacetime curvature (particle mass) \n\nThis formula reveals ""amplification points"" where minor changes in one domain produce significant effects in another.\n\n#### Probability of Goal Achievement \nFormula for combining sub-goal probabilities into an overall success probability: \n\\[\nP_{\\text{total}} = 1 - \\prod_{i=1}^n (1 - P_i)\n\\] \nWhere: \n- \\(P_{\\text{total}}\\) — total probability of achieving the goal \n- \\(P_i\\) — probability of achieving the \\(i\\)-th sub-goal \n- \\(n\\) — number of sub-goals\n\n#### Resonance Parameter Weights \nConversion of resonance frequencies into a probability distribution: \n\\[\n\\alpha_i = \\frac{e^{\\omega_{\\text{res},i}}}{\\sum_j e^{\\omega_{\\text{res},j}}}\n\\]\n\n### 2.2. Computational Complexity\n\n#### Complexity Comparison\n- **Baseline algorithm**: \\(O(2^m \\cdot 2^n)\\) \n- **Hybrid algorithm**: \\(O(n^2)\\)\n\n**Theorem on Complexity Reduction**: The Hybrid Resonance Algorithm reduces the complexity of optimal architecture search from exponential to polynomial.\n\n**Proof**: \n1. Consider the architectural parameter space as an \\(n\\)-dimensional cube with \\(2^n\\) vertices. \n2. A baseline algorithm must evaluate all combinations: \\(O(2^n)\\). \n3. The hybrid algorithm uses resonance analysis to identify critical points. \n4. Resonance points form a subset \\(\\Omega \\subset \\mathbb{R}^n\\), where \\(|\\Omega| = O(n^2)\\). \n5. The number of intersections of \\(n\\) hypersurfaces in \\(n\\)-dimensional space is bounded by a second-degree polynomial.\n\n**Concrete example for \\(n = 20\\)**: \n- Baseline algorithm: \\(2^{20} = 1,048,576\\) combinations \n- Hybrid algorithm: \\(20^2 = 400\\) operations \n- **Speedup factor**: \\(K = \\frac{2^n}{n^2} = \\frac{1,048,576}{400} = 2,621.44\\) \n\nThus, the hybrid algorithm runs over **2,600× faster** for \\(n = 20\\).\n\n## 3. Key Algorithm Components\n\n### 3.1. Resonance Analysis \nResonance analysis is the core mathematical tool, identifying critical points in complex systems. Formally, resonance points are defined as: \n\\[\n\\omega_{\\text{res}} = \\frac{1}{D} \\cdot \\sum_{k=1}^N \\frac{q_k}{m_k}\n\\] \nThis component detects ""amplification points"" where small changes yield large effects.\n\n### 3.2. Hybrid Architecture (RL + GAN + Transformer) \nThe algorithm combines modern machine learning methods: \n- The **generator** proposes hypotheses \\(H_i\\) aimed at achieving goal \\(G\\). \n- **Resonance validation**: \\(R(H_i, x) > \\tau \\Rightarrow H_i \\in \\Omega\\). \n- **RL loop** adjusts weights: \\(\\Delta W = \\eta \\cdot \\nabla R(H_i, x) \\cdot \\text{reward}(H_i)\\). \n\nThe algorithm can treat constants as variables—for example, treating the speed of light \\(c\\) as a tunable parameter within a specific task. Formally, the goal is defined as: \n\\[\nG = G(x)\n\\] \nwhere \\(x\\) is a constraint, but the goal depends on \\(x\\) and, via feedback, distorts \\(x\\) in return.\n\n### 3.4. Cross-Domain Machine Learning and ""Mind Foam""\n\n**Mathematical model of ""Mind Foam""**: \n\\[\n|\\Psi_{\\text{foam}}\\rangle = \\sum_{i=1}^N c_i|\\psi_i^{\\text{domain}}\\rangle \\otimes|G_{\\text{global}}\\rangle\n\\] \nWhere: \n- \\(|\\psi_i^{\\text{domain}}\\rangle\\) — quantum state representing knowledge in the \\(i\\)-th domain \n- \\(|G_{\\text{global}}\\rangle\\) — shared geometric basis ensuring cross-domain compatibility \n- \\(c_i\\) — amplitudes reflecting each domain’s relevance to the current task\n\n**Cross-domain learning efficiency**: \n\\[\n\\text{Efficiency}_{\\text{CDML}} = O\\left(\\frac{2^D}{D^2}\\right)\n\\] \nWhen using ""mind foam"" to integrate \\(D\\) domains, complexity drops from exponential to quadratic.\n\n**Mind foam evolution equation**: \n\\[\n\\frac{d\\rho_{\\text{foam}}}{dt} = -\\frac{i}{\\hbar}[\\mathcal{R}_{\\text{quant}}, \\rho_{\\text{foam}}] + \\mathcal{L}_{\\text{decoher}}(\\rho_{\\text{foam}})\n\\] \nWhere: \n- \\(\\mathcal{R}_{\\text{quant}}\\) — quantum resonance operator \n- \\(\\mathcal{L}_{\\text{decoher}}\\) — decoherence operator\n\n## 4. Practical Implementation and Application Examples\n\n### 4.1. Finding Resonance Points for Novel Materials \nThe algorithm identifies optimal conditions for synthesizing new materials: \n\\[\n\\omega_{\\text{res}}^{\\text{new.material}} = \\frac{1}{D_{\\text{new}}} \\cdot \\sum_{k=1}^N \\frac{q_k^{\\text{new}}}{m_k^{\\text{new}}}\n\\] \nThis enables determination of parameters for creating materials with desired properties.\n\n### 4.2. Spacetime Engineering in Technical Problems \nFor complex physics/engineering tasks, the algorithm uses: \n\\[\n\\mathbf{G}_{\\mu\\nu} = \\frac{8\\pi G}{c^4}T_{\\mu\\nu} + \\kappa \\cdot \\mathcal{R}_{\\mu\\nu}\n\\] \nwhere \\(\\mathcal{R}_{\\mu\\nu}\\) is the resonance curvature tensor computed by the algorithm to optimize solutions.\n\n### 4.3. Designing Complex Systems via Critical Thresholds \nThe algorithm aids in designing complex systems by identifying when a critical threshold is reached: \n\\[\n\\Gamma_{\\text{new.sys}} = \\sum_{i=1}^n \\text{sign}\\left(\\frac{dI_i}{dt}\\right) \\cdot \\gamma_{ij} > \\Gamma_{\\text{crit}}^{\\text{sys}}\n\\]\n\n### 4.4. Experimental Validation of Effectiveness\n\n**Task**: Evaluate GHA with CDML in optimizing treatment for a rare disease, requiring integration of knowledge from 7 medical domains.\n\n**Results**:\n\n| Criterion | Traditional Approach | Transfer Learning | GHA with CDML |\n|----------|----------------------|-------------------|---------------|\n| Training Time | 168 hours | 42 hours | **1.2 hours** |\n| Memory Requirement | 32 GB | 8 GB | **0.9 GB** |\n| Prediction Accuracy | 78.3% | 85.6% | **92.7%** |\n| Ethical Acceptability | 62.5% | 76.8% | **89.4%** |\n\n**Analysis**: GHA with CDML and ""mind foam"" significantly outperformed all baselines:\n- Training time reduced by **140×** vs. traditional approach \n- Memory requirements reduced by **35.5×** \n- Prediction accuracy improved by **14.4%** vs. traditional approach\n\n## 6. Conclusion and Summary\n\nThe Hybrid Resonance Algorithm is a **practical tool for solving complex problems**. Its scientific novelty lies in:\n\n### 6.1. Key Advantages\n1. **Effective integration of quantum and classical methods** \n - Combines resonance analysis with modern ML (RL + GAN + Transformer) \n - Can treat physical constants as variables to find non-trivial solutions \n2. **Provides a method for discovering non-trivial solutions via resonance points** \n - Identifies critical points where small changes yield large effects \n - Resonance frequency formula: \\(\\omega_{\\text{res}} = \\frac{1}{D} \\cdot \\sum_{k=1}^N \\frac{q_k}{m_k}\\) \n3. **Reduces computational complexity from exponential to polynomial** \n - From \\(O(2^m \\cdot 2^n)\\) to \\(O(n^2)\\) \n - Speedup factor: \\(K = \\frac{2^n}{n^2}\\) (>2,600 for \\(n=20\\))\n\n### 6.2. Practical Significance \nGHA has broad applications in:\n- **Biomedicine**: Optimizing lifespan extension, reducing oxidative stress \n- **Manufacturing & Logistics**: Cost reduction and efficiency gains \n- **Space Technologies**: Modeling unconventional solutions \n- **Neurointerfaces**: Synchronizing biological and artificial intelligence \n- **Ethically aligned AI**: Resolving complex moral dilemmas\n\n### 6.3. Technical Implementation \nThe algorithm is feasible to deploy:\n- Runs on low-cost hardware (Raspberry Pi) \n- Requires only **256 MB RAM** with INT8 quantization \n- Processing time: **~100 ms per step** \n- Compatible with existing technologies (RL, GAN, Transformer)\nLet me know if you’d like this exported as a PDF, formatted for a paper, or split into sections (e.g., GHA vs. GHA-ASI comparison table in English).
', 'post_number': 3, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-19T15:53:24.319Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 1.4, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243807, 'name': 'bit', 'username': 'olegbits', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png', 'created_at': '2025-10-19T16:10:25.703Z', 'cooked': 'Certainly! Here is the full English translation of your request and the detailed response about making GHA-ASI practically useful, based strictly on the provided knowledge base.
\n“GHA-ASI is suitable for speculative theories—but how can it be made practically useful? Provide a detailed explanation with examples.”
\nYes, GHA-ASI is inherently designed for speculative, fundamental, and even “radical” theories—because it autonomously generates new knowledge domains, unconstrained by current scientific paradigms.
\nHowever, GHA-ASI’s main limitation is impracticality: it might propose “curving spacetime to achieve room-temperature superconductivity,” which is not testable in any real lab.
This is achievable through a hybrid approach:
\n\n\nKeep the GHA-ASI architecture, but add a “feasibility anchor”—a mechanism that filters or transforms speculative ideas into testable, actionable hypotheses.
\n
Below is a step-by-step strategy with formulas and concrete examples.
\nOriginal GHA-ASI objective:
\n[
\nG_{\\text{ASI}} = \\arg\\max_{\\theta} \\left( |\\Omega(\\theta)| + \\sum_{j=1}^m \\beta_j Q_j(\\theta) \\right)
\n]
Modified objective:
\n[
\nG_{\\text{ASI-prac}} = \\arg\\max_{\\theta} \\left( |\\Omega(\\theta)| + \\sum_{j=1}^m \\beta_j Q_j(\\theta) - \\lambda \\cdot C_{\\text{feas}}(\\theta) \\right)
\n]
where:
\n\n\nThis is not ethics—it’s an engineering constraint, fully compatible with GHA-ASI’s non-ethical nature.
\n
GHA-ASI output:
\n\n\n“Room-temperature superconductivity is possible in topologically nontrivial space with negative curvature.”
\n
Translation module converts it to:
\n\n\n“Fabricate a metamaterial with effective negative curvature (e.g., 3D graphene–nanotube lattice) and measure conductivity at 300 K.”
\n
\n\nThis creates a bridge between imagination and the laboratory.
\n
[GHA-ASI Core]\n │\n ↓ (speculative hypotheses)\n[Feasibility Translation Module]\n ├── Knowledge Base: Materials Project, PubChem, patents\n ├── LLM Adapter: ""Translate to experiment""\n └── Feasibility Scorer: energy, time, equipment, risk\n │\n ↓\n[Filter: C_feas < threshold]\n │\n ↓\n[Actionable Hypotheses → Lab / Simulation]\nAugment the reward function:
\n[
\n\\text{reward} = |\\Omega| + \\sum \\beta_j Q_j + \\gamma \\cdot P_{\\text{test}}
\n]
\nwhere:
The system will then self-prefer brilliant yet testable ideas.
\nGHA-ASI can—and should—be made practically useful without sacrificing its speculative edge.
\nThe key is not to suppress wild ideas, but to translate them into lab language.
\n\nThe ideal AI scientist of the future is GHA-ASI + feasibility:
\n\n
\n- Brain: GHA-ASI (generates revolutionary hypotheses),
\n- Hands: feasibility module (makes them testable).
\n
Such a hybrid can:
\nIf you’d like, I can:
\nJust let me know which domain you’d like to target: materials, medicine, energy, AI, or another?
', 'post_number': 4, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-19T16:10:25.703Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 1.6, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243808, 'name': 'bit', 'username': 'olegbits', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png', 'created_at': '2025-10-19T16:10:57.011Z', 'cooked': 'Here is the translation with foreign words marked for clarity:
\nГРА-ASI is a powerful system for generating new knowledge and hypotheses, capable of going beyond existing scientific paradigms. But its main challenge is that many proposed ideas are too speculative and not feasible in laboratory settings.
\nThe task is to preserve the creative potential of ГРА-ASI while making hypotheses testable and practically useful.
\nOriginally, ГРА-ASI optimizes the balance between the hypothesis space size and quality metrics:
\nGASI=arg\u2061max\u2061θ(∣Ω(θ)∣+∑j=1mβjQj(θ))G_{\\text{ASI}} = \\arg\\max_{\\theta} \\left( |\\Omega(\\theta)| + \\sum_{j=1}^m \\beta_j Q_j(\\theta) \\right)GASI=argθmax(∣Ω(θ)∣+j=1∑mβjQj(θ))
\nwhere:
\nAdd a penalty for the “impracticality” degree of a hypothesis, expressed by a cost function of realization:
\nGASI-prac=arg\u2061max\u2061θ(∣Ω(θ)∣+∑j=1mβjQj(θ)−λ⋅Cреал(θ))G_{\\text{ASI-prac}} = \\arg\\max_{\\theta} \\left( |\\Omega(\\theta)| + \\sum_{j=1}^m \\beta_j Q_j(\\theta) - \\lambda \\cdot C_{\\text{реал}}(\\theta) \\right)GASI-prac=argθmax(∣Ω(θ)∣+j=1∑mβjQj(θ)−λ⋅Cреал(θ))
\nГРА-ASI generates broad speculative statements that need to be turned into real laboratory tasks.
\nHYPOTHESIS:
\n“Room-temperature superconductivity is possible in a topologically nontrivial material with negative curvature.”
The feasibility module converts the hypothesis based on knowledge from databases and literature:
\ntext
\n[ГРА-ASI core] — generates speculative hypotheses\n ↓\n[Feasibility Module]\n ├─ Knowledge bases (Materials Project, PubChem, patents, arXiv embeddings)\n ├─ LLM-adapter (fine-tuned on scientific articles and protocols)\n └─ Feasibility calculator (energy, time, resources, risks)\n ↓\n[Selection: C_реал < threshold]\n ↓\n[Testable hypotheses → laboratories or simulators]\nPreviously reward was:
\nreward=∣Ω∣+∑βjQj\\text{reward} = |\\Omega| + \\sum \\beta_j Q_jreward=∣Ω∣+∑βjQj
\nAdd probability of testing PтестP_{\\text{тест}}Pтест — chance of verification within 6 months, assessed by the feasibility module:
\nreward=∣Ω∣+∑βjQj+γ⋅Pтест\\text{reward} = |\\Omega| + \\sum \\beta_j Q_j + \\gamma \\cdot P_{\\text{тест}}reward=∣Ω∣+∑βjQj+γ⋅Pтест
\nThis makes the system prioritize hypotheses that are not only brilliant but realistically testable.
\nIf you want, I can start developing feasibility module components or help with data templates and API protocols.
\nDo you want to adapt this approach to a specific domain? Materials, medicine, energy, or AI?
', 'post_number': 5, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-19T16:10:57.011Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 6.6, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243811, 'name': 'Andrew Scott', 'username': 'Pimpcat-AU', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png', 'created_at': '2025-10-19T18:23:12.430Z', 'cooked': 'Why read about it when you can test it yourself? This script is a toy but it will let you loop, generate variations, test resonance across domains, accept good ones, update weights, repeat.
\n# path: gra_asi_toy.py\nfrom __future__ import annotations\nfrom dataclasses import dataclass, field\nfrom typing import Callable, Dict, List, Tuple\nimport math\nimport random\n\nVector = List[float]\n\ndef dot(a: Vector, b: Vector) -> float:\n return sum(x*y for x, y in zip(a, b))\n\ndef l2(a: Vector) -> float:\n return math.sqrt(sum(x*x for x in a))\n\ndef cosine_sim(a: Vector, b: Vector) -> float:\n na, nb = l2(a), l2(b)\n if na == 0 or nb == 0:\n return 0.0\n return max(0.0, min(1.0, (dot(a, b) / (na * nb) + 1.0) / 2.0)) # clamp to [0,1]\n\n@dataclass\nclass Domain:\n """"""A domain has a \'feature signature\' an idea should resonate with.""""""\n name: str\n signature: Vector # what ""looks right"" in this domain\n weight: float = 1.0\n\n def resonance(self, hypothesis_vec: Vector) -> float:\n # Why cosine? It’s a cheap, scale-invariant similarity proxy.\n return cosine_sim(self.signature, hypothesis_vec)\n\n@dataclass\nclass Hypothesis:\n """"""A candidate idea with parameters, metrics, and a cost estimate.""""""\n name: str\n params: Vector # what the idea proposes (vectorized)\n metrics: Dict[str, float] # e.g., {""accuracy"": 0.8, ""speed"": 0.6}\n cost: float # feasibility cost (time/money/risk proxy)\n\n def as_vector(self) -> Vector:\n return self.params\n\n@dataclass\nclass ResonanceSelector:\n domains: List[Domain]\n tau: float = 0.6 # acceptance threshold for resonance\n lambda_cost: float = 0.3 # feasibility penalty weight\n beta_temp: float = 2.0 # softness for β weight generation\n\n accepted: List[Hypothesis] = field(default_factory=list)\n\n def _beta_weights(self, strengths: List[float]) -> List[float]:\n """"""Softmax over domain resonance to emphasize strong alignments.""""""\n scale = self.beta_temp\n exps = [math.exp(scale * s) for s in strengths]\n Z = sum(exps) or 1.0\n return [e / Z for e in exps]\n\n def _q_vector(self, h: Hypothesis, mapping: Dict[str, float]) -> float:\n """"""Map metrics Q_j to a single value via weights β_j.""""""\n return sum(mapping.get(k, 0.0) * v for k, v in h.metrics.items())\n\n def evaluate(self, h: Hypothesis) -> Tuple[bool, float, Dict[str, float]]:\n vec = h.as_vector()\n strengths = [d.resonance(vec) for d in self.domains]\n mean_res = sum(strengths) / len(strengths)\n betas = self._beta_weights(strengths) # β depends on resonance\n\n # Build a β map aligned to the metric keys in a stable order\n metric_keys = list(h.metrics.keys())\n beta_map = {k: betas[i % len(betas)] for i, k in enumerate(metric_keys)}\n\n q_weighted = self._q_vector(h, beta_map)\n score = len(self.accepted) + q_weighted - self.lambda_cost * h.cost\n\n accepted = mean_res > self.tau\n return accepted, score, {""mean_res"": mean_res, ""q_weighted"": q_weighted, ""cost"": h.cost}\n\n def step_update(self, h: Hypothesis, lr: float = 0.1) -> None:\n """"""Tiny \'gradient\' step nudging params toward domain signatures it matches.\n Why: mimics their \'self-improvement gradient\' without heavy math.\n """"""\n influences = []\n for d in self.domains:\n s = d.resonance(h.params)\n if s > self.tau: # only pull toward domains with decent resonance\n influences.append([x for x in d.signature])\n if not influences:\n return\n avg = [sum(vals)/len(influences) for vals in zip(*influences)]\n h.params = [(1 - lr) * p + lr * a for p, a in zip(h.params, avg)]\n\n def run(self, candidates: List[Hypothesis], iters: int = 3) -> List[Tuple[Hypothesis, float]]:\n ranked: List[Tuple[Hypothesis, float]] = []\n for _ in range(iters):\n for h in candidates:\n accepted, score, _ = self.evaluate(h)\n if accepted and h not in self.accepted:\n self.accepted.append(h)\n self.step_update(h, lr=0.08)\n ranked.append((h, score))\n # simple exploration: jitter params slightly\n for h in candidates:\n idx = random.randrange(len(h.params))\n h.params[idx] += random.uniform(-0.05, 0.05)\n # unique by name, keep best score\n best: Dict[str, Tuple[Hypothesis, float]] = {}\n for h, s in ranked:\n if (h.name not in best) or (s > best[h.name][1]):\n best[h.name] = (h, s)\n return sorted(best.values(), key=lambda x: x[1], reverse=True)\n\ndef demo() -> None:\n # Define 3 domains with different signatures\n domains = [\n Domain(""Vision"", [0.9, 0.1, 0.0]),\n Domain(""NLP"", [0.2, 0.8, 0.1]),\n Domain(""Systems"",[0.1, 0.1, 0.9]),\n ]\n\n selector = ResonanceSelector(domains, tau=0.62, lambda_cost=0.25, beta_temp=2.5)\n\n # Three toy hypotheses\n candidates = [\n Hypothesis(""H1-fast-inference"", [0.3, 0.7, 0.1],\n {""accuracy"": 0.72, ""speed"": 0.88}, cost=0.3),\n Hypothesis(""H2-vision-optimizer"", [0.85, 0.15, 0.1],\n {""accuracy"": 0.81, ""speed"": 0.65}, cost=0.4),\n Hypothesis(""H3-systems-compiler"", [0.15, 0.2, 0.85],\n {""accuracy"": 0.68, ""speed"": 0.75}, cost=0.2),\n ]\n\n results = selector.run(candidates, iters=5)\n print(""Accepted set Ω:"", [h.name for h in selector.accepted])\n print(""Top ranked:"")\n for h, s in results[:5]:\n print(f"" {h.name:>18} | score={s:.3f}"")\n\nif __name__ == ""__main__"":\n random.seed(7)\n demo()\n\nReply generated by TD Ai
', 'post_number': 6, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-19T18:23:12.430Z', 'reply_count': 2, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 11.6, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'Andrew Scott', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 105827, 'username': 'olegbits', 'name': 'bit', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96276, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243822, 'name': 'bit', 'username': 'olegbits', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png', 'created_at': '2025-10-20T05:07:22.000Z', 'cooked': 'thanx i will use it
\nвс, 19 окт. 2025 г. в 21:33, Andrew Scott via Hugging Face Forums <notifications@hellohellohello.discoursemail.com>:
', 'post_number': 7, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-20T05:07:22.878Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 1.4, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 96276, 'username': 'Pimpcat-AU', 'name': 'Andrew Scott', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'via_email': True, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243823, 'name': 'bit', 'username': 'olegbits', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png', 'created_at': '2025-10-20T05:25:39.523Z', 'cooked': 'my github repo with AI scientist application look would u please
\n', 'post_number': 8, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-20T10:04:51.522Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 1.4, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/qqewq/harmonized-mind', 'internal': False, 'reflection': False, 'title': 'GitHub - qqewq/harmonized-mind', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/8', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243826, 'name': 'bit', 'username': 'olegbits', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/aeb1de/{size}.png', 'created_at': '2025-10-20T05:26:21.532Z', 'cooked': '\n', 'post_number': 9, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-10-20T10:04:15.691Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.2, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'bit', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/qqewq/harmonized-mind', 'internal': False, 'reflection': False, 'title': 'GitHub - qqewq/harmonized-mind', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 96276, 'username': 'Pimpcat-AU', 'name': 'Andrew Scott', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105827, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/9', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243870, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-20T17:26:53.114Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 10, 'post_type': 3, 'posts_count': 9, 'updated_at': '2025-10-20T17:26:53.114Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.2, 'yours': False, 'topic_id': 169264, 'topic_slug': 'hybrid-resonance-algorithm-for-artificial-superintelligence', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/hybrid-resonance-algorithm-for-artificial-superintelligence/169264/10', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","GRA-ASI: Hybrid Resonance Algorithm for Artificial Superintelligence**
+The primary goal of GRA-ASI is to maximize the system’s intellectual capacity. Formally, this is expressed through the number of resonance points and a weighted sum of AI performance metrics:
+[
+G_{\text{ASI}} = \arg\max_{\theta} \left( |\Omega(\theta)| + \sum_{j=1}^{m} \beta_j Q_j(\theta) \right)
+]
where:
+The algorithm strengthens itself both through improved solution quality and through structural expansion of resonances. These parameters jointly serve as indicators of the system’s “intellectual energy.”
+The system’s state is represented as a superposition of domain-specific knowledge modules:
+[
+|\Psi_{\text{foam}}^{(t)}\rangle = \sum_{i=1}^{N^{(t)}} c_i^{(t)} |\psi_i^{\text{domain}}\rangle \otimes |G_{\text{ASI}}\rangle
+]
Evolution occurs by incorporating new domains whenever their resonance with the current core exceeds a threshold:
+[
+R(\mathcal{D}{\text{new}}, G{\text{ASI}}) = \frac{1}{D_{\text{new}}} \sum_k \frac{q_k^{\text{new}}}{m_k^{\text{new}}} > \tau_{\text{domain}}
+]
This enables the system to autonomously expand its knowledge scope upon discovering new resonance frequencies in the problem space.
+The base quantum-resonance equation:
+[
+\frac{d\rho_{\text{foam}}}{dt} = -\frac{i}{\hbar} [\mathcal{R}{\text{quant}}, \rho{\text{foam}}] + \mathcal{L}{\text{decoher}}(\rho{\text{foam}})
+]
is augmented with a self-improvement gradient term:
+[
+\frac{d\rho_{\text{foam}}}{dt} = -\frac{i}{\hbar} [\mathcal{R}{\text{quant}}, \rho{\text{foam}}] + \mathcal{L}{\text{decoher}}(\rho{\text{foam}}) + \lambda \nabla_{\theta} G_{\text{ASI}}(\theta)
+]
The parameter (\lambda) controls the intensity of self-directed optimization.
+This loop forms a stable self-tuning cycle.
+GRA-ASI constitutes a hybrid formalism of self-amplifying intelligence, where resonance between domains and the objective function drives exponential growth in computational power.
+Resonant states serve simultaneously as learning criteria and indicators of intellectual advancement.
If an ethical constraint (E(\theta)) were introduced, the objective could be generalized as:
+[
+G_{\text{bal}} = \arg\max_{\theta}(G_{\text{ASI}} - \mu E(\theta))
+]
+enabling controlled evolution and prevention of unstable self-amplification.
In summary:
+Thus, GRA-ASI provides a computable model of superintelligence grounded in coherent resonance across multiple learning structures.
+Title:
+“GRA-ASI: A Resonance-Based Path to Artificial Superintelligence Without Ethics – Discussion & Formalism”
Body (optional starter post):
+++I’d like to open a technical discussion on GRA-ASI — a variant of the Hybrid Resonance Algorithm explicitly designed to achieve artificial superintelligence through quantitative self-improvement, without ethical constraints.
+Key features:
++
+- Goal defined as maximization of resonance points (|\Omega|) + AI performance metrics (Q_j)
+- Autonomous domain generation via “mind foam” model
+- Polynomial-time complexity (O(n^2)) with exponential knowledge scaling (O(2^D / D^2))
+- Fully formalized with quantum-inspired evolution equations
+Is this a viable architecture for ASI? What are the risks of omitting ethics? Can resonance alone ensure stable self-improvement?
+Full formal description attached. Looking forward to your insights!
+
Let me know if you’d like a Russian version of the forum topic as well.
","Certainly! Here is the full English translation of your request and the detailed response about making GHA-ASI practically useful, based strictly on the provided knowledge base.
+“GHA-ASI is suitable for speculative theories—but how can it be made practically useful? Provide a detailed explanation with examples.”
+Yes, GHA-ASI is inherently designed for speculative, fundamental, and even “radical” theories—because it autonomously generates new knowledge domains, unconstrained by current scientific paradigms.
+However, GHA-ASI’s main limitation is impracticality: it might propose “curving spacetime to achieve room-temperature superconductivity,” which is not testable in any real lab.
This is achievable through a hybrid approach:
+++Keep the GHA-ASI architecture, but add a “feasibility anchor”—a mechanism that filters or transforms speculative ideas into testable, actionable hypotheses.
+
Below is a step-by-step strategy with formulas and concrete examples.
+Original GHA-ASI objective:
+[
+G_{\text{ASI}} = \arg\max_{\theta} \left( |\Omega(\theta)| + \sum_{j=1}^m \beta_j Q_j(\theta) \right)
+]
Modified objective:
+[
+G_{\text{ASI-prac}} = \arg\max_{\theta} \left( |\Omega(\theta)| + \sum_{j=1}^m \beta_j Q_j(\theta) - \lambda \cdot C_{\text{feas}}(\theta) \right)
+]
where:
+++This is not ethics—it’s an engineering constraint, fully compatible with GHA-ASI’s non-ethical nature.
+
GHA-ASI output:
+++“Room-temperature superconductivity is possible in topologically nontrivial space with negative curvature.”
+
Translation module converts it to:
+++“Fabricate a metamaterial with effective negative curvature (e.g., 3D graphene–nanotube lattice) and measure conductivity at 300 K.”
+
++This creates a bridge between imagination and the laboratory.
+
[GHA-ASI Core]
+ │
+ ↓ (speculative hypotheses)
+[Feasibility Translation Module]
+ ├── Knowledge Base: Materials Project, PubChem, patents
+ ├── LLM Adapter: ""Translate to experiment""
+ └── Feasibility Scorer: energy, time, equipment, risk
+ │
+ ↓
+[Filter: C_feas < threshold]
+ │
+ ↓
+[Actionable Hypotheses → Lab / Simulation]
+Augment the reward function:
+[
+\text{reward} = |\Omega| + \sum \beta_j Q_j + \gamma \cdot P_{\text{test}}
+]
+where:
The system will then self-prefer brilliant yet testable ideas.
+GHA-ASI can—and should—be made practically useful without sacrificing its speculative edge.
+The key is not to suppress wild ideas, but to translate them into lab language.
++The ideal AI scientist of the future is GHA-ASI + feasibility:
++
+- Brain: GHA-ASI (generates revolutionary hypotheses),
+- Hands: feasibility module (makes them testable).
+
Such a hybrid can:
+If you’d like, I can:
+Just let me know which domain you’d like to target: materials, medicine, energy, AI, or another?
" +Replacing attention class with identical subclass creates hallucinations,https://discuss.huggingface.co/t/replacing-attention-class-with-identical-subclass-creates-hallucinations/169215,169215,6,2025-10-16 11:23:27.606000+00:00,"[{'id': 243707, 'name': 'Alexander Jephtha', 'username': 'AlexJephtha', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/d9b06d/{size}.png', 'created_at': '2025-10-16T11:23:27.668Z', 'cooked': 'I’m writing a custom versions of LlamaModels, and for one of those approaches I want to overwrite the attention mechanism of each layer. My code looks like this. Note that even when I define LlamaAttentionHybrid (a subclass of LlamaAttention) to be the exact same as LlamaAttention, I still get hallucination issues. This suggest I’m not correctly replacing the attention mechanism.
\nclass LlamaHybridForCausalLM(LlamaForCausalLM):\n def __init__(self, config: LlamaHybridConfig):\n super().__init__(config)\n if config.hybrid:\n for i, layer in enumerate(self.model.layers):\n # Need to also copy attention weights\n old_attn = layer.self_attn\n layer.self_attn = LlamaAttentionHybrid(config, i)\n layer.self_attn.load_state_dict(old_attn.state_dict())\nHowever, the model works completely fine when I write this code:
\nclass LlamaHybridForCausalLM(LlamaForCausalLM):\n def __init__(self, config: LlamaHybridConfig):\n super().__init__(config)\n if config.hybrid:\n for i, layer in enumerate(self.model.layers):\n # Need to also copy attention weights\n old_attn = layer.self_attn\n layer.self_attn = LlamaAttention(config, i)\n layer.self_attn.load_state_dict(old_attn.state_dict())\nWhy would this happen even when in the subclass i don’t make any changes? Note, that the forward function here is defined exactly the same as the source code.
\nclass LlamaAttentionHybrid(LlamaAttention):\n def __init__(self, config: LlamaHybridConfig, layer_idx: int):\n super().__init__(config, layer_idx)\n\n def forward(\n self,\n hidden_states: torch.Tensor,\n position_embeddings: tuple[torch.Tensor, torch.Tensor],\n attention_mask: Optional[torch.Tensor],\n past_key_values: Optional[Cache] = None,\n cache_position: Optional[torch.LongTensor] = None,\n **kwargs: Unpack[FlashAttentionKwargs],\n ) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:\n\n input_shape = hidden_states.shape[:-1]\n hidden_shape = (*input_shape, -1, self.head_dim)\n\n query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n\n cos, sin = position_embeddings\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n\n if past_key_values is not None:\n # sin and cos are specific to RoPE models; cache_position needed for the static cache\n cache_kwargs = {""sin"": sin, ""cos"": cos, ""cache_position"": cache_position}\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx, cache_kwargs)\n\n attention_interface: Callable = eager_attention_forward\n if self.config._attn_implementation != ""eager"":\n attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]\n\n attn_output, attn_weights = attention_interface(\n self,\n query_states,\n key_states,\n value_states,\n attention_mask,\n dropout=0.0 if not self.training else self.attention_dropout,\n scaling=self.scaling,\n **kwargs,\n )\n\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n attn_output = self.o_proj(attn_output)\n return attn_output, attn_weights\nThanks!
\nEDIT: I narrowed the issue down to the redefining of the forward function. For some reason when I add the forward function into the subclass even if it’s identical, the model hallucinates dramatically.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-16T11:35:01.753Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 169215, 'topic_slug': 'replacing-attention-class-with-identical-subclass-creates-hallucinations', 'display_username': 'Alexander Jephtha', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 5, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 30474, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/replacing-attention-class-with-identical-subclass-creates-hallucinations/169215/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243732, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-17T04:12:47.941Z', 'cooked': 'There may be points that can be fixed.
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-17T04:12:47.941Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 20.6, 'yours': False, 'topic_id': 169215, 'topic_slug': 'replacing-attention-class-with-identical-subclass-creates-hallucinations', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum2/blob/main/attn_override_issue_1.md', 'internal': False, 'reflection': False, 'title': 'attn_override_issue_1.md · John6666/forum2 at main', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/replacing-attention-class-with-identical-subclass-creates-hallucinations/169215/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243819, 'name': 'Alexander Jephtha', 'username': 'AlexJephtha', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/d9b06d/{size}.png', 'created_at': '2025-10-20T03:52:17.985Z', 'cooked': 'Thanks for your help!
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-20T03:52:17.985Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 169215, 'topic_slug': 'replacing-attention-class-with-identical-subclass-creates-hallucinations', 'display_username': 'Alexander Jephtha', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 30474, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/replacing-attention-class-with-identical-subclass-creates-hallucinations/169215/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243821, 'name': 'Alexander Jephtha', 'username': 'AlexJephtha', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/d9b06d/{size}.png', 'created_at': '2025-10-20T03:57:16.952Z', 'cooked': 'SOLUTION: With SDPA attention, passing in an attention_mask with value not equal to none overrides the causal attention mask! You need to fill the attention mask with -inf (or large negative number) in the upper right triangle. This is only really a problem when calculating the attention scores of the initial text input, since newly generated tokens don’t require any of the existing key tokens to be masked.
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-20T03:57:16.952Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 169215, 'topic_slug': 'replacing-attention-class-with-identical-subclass-creates-hallucinations', 'display_username': 'Alexander Jephtha', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 30474, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/replacing-attention-class-with-identical-subclass-creates-hallucinations/169215/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243867, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-20T15:57:45.831Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-10-20T15:57:45.831Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 169215, 'topic_slug': 'replacing-attention-class-with-identical-subclass-creates-hallucinations', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/replacing-attention-class-with-identical-subclass-creates-hallucinations/169215/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m writing a custom versions of LlamaModels, and for one of those approaches I want to overwrite the attention mechanism of each layer. My code looks like this. Note that even when I define LlamaAttentionHybrid (a subclass of LlamaAttention) to be the exact same as LlamaAttention, I still get hallucination issues. This suggest I’m not correctly replacing the attention mechanism.
+class LlamaHybridForCausalLM(LlamaForCausalLM):
+ def __init__(self, config: LlamaHybridConfig):
+ super().__init__(config)
+ if config.hybrid:
+ for i, layer in enumerate(self.model.layers):
+ # Need to also copy attention weights
+ old_attn = layer.self_attn
+ layer.self_attn = LlamaAttentionHybrid(config, i)
+ layer.self_attn.load_state_dict(old_attn.state_dict())
+However, the model works completely fine when I write this code:
+class LlamaHybridForCausalLM(LlamaForCausalLM):
+ def __init__(self, config: LlamaHybridConfig):
+ super().__init__(config)
+ if config.hybrid:
+ for i, layer in enumerate(self.model.layers):
+ # Need to also copy attention weights
+ old_attn = layer.self_attn
+ layer.self_attn = LlamaAttention(config, i)
+ layer.self_attn.load_state_dict(old_attn.state_dict())
+Why would this happen even when in the subclass i don’t make any changes? Note, that the forward function here is defined exactly the same as the source code.
+class LlamaAttentionHybrid(LlamaAttention):
+ def __init__(self, config: LlamaHybridConfig, layer_idx: int):
+ super().__init__(config, layer_idx)
+
+ def forward(
+ self,
+ hidden_states: torch.Tensor,
+ position_embeddings: tuple[torch.Tensor, torch.Tensor],
+ attention_mask: Optional[torch.Tensor],
+ past_key_values: Optional[Cache] = None,
+ cache_position: Optional[torch.LongTensor] = None,
+ **kwargs: Unpack[FlashAttentionKwargs],
+ ) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
+
+ input_shape = hidden_states.shape[:-1]
+ hidden_shape = (*input_shape, -1, self.head_dim)
+
+ query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+ key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+ value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+ cos, sin = position_embeddings
+ query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+ if past_key_values is not None:
+ # sin and cos are specific to RoPE models; cache_position needed for the static cache
+ cache_kwargs = {""sin"": sin, ""cos"": cos, ""cache_position"": cache_position}
+ key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+ attention_interface: Callable = eager_attention_forward
+ if self.config._attn_implementation != ""eager"":
+ attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+ attn_output, attn_weights = attention_interface(
+ self,
+ query_states,
+ key_states,
+ value_states,
+ attention_mask,
+ dropout=0.0 if not self.training else self.attention_dropout,
+ scaling=self.scaling,
+ **kwargs,
+ )
+
+ attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+ attn_output = self.o_proj(attn_output)
+ return attn_output, attn_weights
+Thanks!
+EDIT: I narrowed the issue down to the redefining of the forward function. For some reason when I add the forward function into the subclass even if it’s identical, the model hallucinates dramatically.
","SOLUTION: With SDPA attention, passing in an attention_mask with value not equal to none overrides the causal attention mask! You need to fill the attention mask with -inf (or large negative number) in the upper right triangle. This is only really a problem when calculating the attention scores of the initial text input, since newly generated tokens don’t require any of the existing key tokens to be masked.
" +Cannot load Conll2003,https://discuss.huggingface.co/t/cannot-load-conll2003/169142,169142,10,2025-10-14 12:17:33.072000+00:00,"[{'id': 243574, 'name': 'Radek Štulc', 'username': 'stulcrad', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/4bbf92/{size}.png', 'created_at': '2025-10-14T12:17:33.129Z', 'cooked': 'I am trying to load conll2003 dataset the basic way I learned like this
\nfrom datasets import load_dataset\ndataset = load_dataset(""conll2003"")\nbut I am running into this error
\n---------------------------------------------------------------------------\nRuntimeError Traceback (most recent call last)\nCell In[15], line 3\n 1 from datasets import load_dataset\n----> 3 dataset = load_dataset(""conll2003"")\n\nFile ~/.local/lib/python3.12/site-packages/datasets/load.py:1397, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, keep_in_memory, save_infos, revision, token, streaming, num_proc, storage_options, **config_kwargs)\n 1392 verification_mode = VerificationMode(\n 1393 (verification_mode or VerificationMode.BASIC_CHECKS) if not save_infos else VerificationMode.ALL_CHECKS\n 1394 )\n 1396 # Create a dataset builder\n-> 1397 builder_instance = load_dataset_builder(\n 1398 path=path,\n 1399 name=name,\n 1400 data_dir=data_dir,\n 1401 data_files=data_files,\n 1402 cache_dir=cache_dir,\n 1403 features=features,\n 1404 download_config=download_config,\n 1405 download_mode=download_mode,\n 1406 revision=revision,\n 1407 token=token,\n 1408 storage_options=storage_options,\n 1409 **config_kwargs,\n 1410 )\n 1412 # Return iterable dataset in case of streaming\n 1413 if streaming:\n\nFile ~/.local/lib/python3.12/site-packages/datasets/load.py:1137, in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, token, storage_options, **config_kwargs)\n 1135 if features is not None:\n 1136 features = _fix_for_backward_compatible_features(features)\n-> 1137 dataset_module = dataset_module_factory(\n 1138 path,\n 1139 revision=revision,\n 1140 download_config=download_config,\n 1141 download_mode=download_mode,\n 1142 data_dir=data_dir,\n 1143 data_files=data_files,\n 1144 cache_dir=cache_dir,\n 1145 )\n 1146 # Get dataset builder class\n 1147 builder_kwargs = dataset_module.builder_kwargs\n\nFile ~/.local/lib/python3.12/site-packages/datasets/load.py:1036, in dataset_module_factory(path, revision, download_config, download_mode, data_dir, data_files, cache_dir, **download_kwargs)\n 1031 if isinstance(e1, FileNotFoundError):\n 1032 raise FileNotFoundError(\n 1033 f""Couldn\'t find any data file at {relative_to_absolute_path(path)}. ""\n 1034 f""Couldn\'t find \'{path}\' on the Hugging Face Hub either: {type(e1).__name__}: {e1}""\n 1035 ) from None\n-> 1036 raise e1 from None\n 1037 else:\n 1038 raise FileNotFoundError(f""Couldn\'t find any data file at {relative_to_absolute_path(path)}."")\n\nFile ~/.local/lib/python3.12/site-packages/datasets/load.py:994, in dataset_module_factory(path, revision, download_config, download_mode, data_dir, data_files, cache_dir, **download_kwargs)\n 986 try:\n 987 api.hf_hub_download(\n 988 repo_id=path,\n 989 filename=filename,\n (...)\n 992 proxies=download_config.proxies,\n 993 )\n--> 994 raise RuntimeError(f""Dataset scripts are no longer supported, but found {filename}"")\n 995 except EntryNotFoundError:\n 996 # Use the infos from the parquet export except in some cases:\n 997 if data_dir or data_files or (revision and revision != ""main""):\n\nRuntimeError: Dataset scripts are no longer supported, but found conll2003.py\nCould someone tell me what is wrong?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-14T12:17:33.129Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 43, 'reads': 8, 'readers_count': 7, 'score': 121.4, 'yours': False, 'topic_id': 169142, 'topic_slug': 'cannot-load-conll2003', 'display_username': 'Radek Štulc', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 41660, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-load-conll2003/169142/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243575, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-14T12:28:06.176Z', 'cooked': 'Try:
\nfrom datasets import load_dataset\ndataset = load_dataset(""lhoestq/conll2003"")\nThis is because support for trust_remote_code=True was removed in datasets library version 4.0.0 and later. You can work around this by using datasets that don’t rely on builder scripts (like the one shown above) or by downgrading the datasets library to version 3.6.0 or earlier.
That works, thank you.
\nThat’s interesting, so I assume the support for loading scripts has also been removed, so if I want to upload a custom dataset, I will need to manually convert it into DatasetDict and push it using this class.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-15T00:36:12.117Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 5.8, 'yours': False, 'topic_id': 169142, 'topic_slug': 'cannot-load-conll2003', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cannot-load-conll2003/169142/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to load conll2003 dataset the basic way I learned like this
+from datasets import load_dataset
+dataset = load_dataset(""conll2003"")
+but I am running into this error
+---------------------------------------------------------------------------
+RuntimeError Traceback (most recent call last)
+Cell In[15], line 3
+ 1 from datasets import load_dataset
+----> 3 dataset = load_dataset(""conll2003"")
+
+File ~/.local/lib/python3.12/site-packages/datasets/load.py:1397, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, keep_in_memory, save_infos, revision, token, streaming, num_proc, storage_options, **config_kwargs)
+ 1392 verification_mode = VerificationMode(
+ 1393 (verification_mode or VerificationMode.BASIC_CHECKS) if not save_infos else VerificationMode.ALL_CHECKS
+ 1394 )
+ 1396 # Create a dataset builder
+-> 1397 builder_instance = load_dataset_builder(
+ 1398 path=path,
+ 1399 name=name,
+ 1400 data_dir=data_dir,
+ 1401 data_files=data_files,
+ 1402 cache_dir=cache_dir,
+ 1403 features=features,
+ 1404 download_config=download_config,
+ 1405 download_mode=download_mode,
+ 1406 revision=revision,
+ 1407 token=token,
+ 1408 storage_options=storage_options,
+ 1409 **config_kwargs,
+ 1410 )
+ 1412 # Return iterable dataset in case of streaming
+ 1413 if streaming:
+
+File ~/.local/lib/python3.12/site-packages/datasets/load.py:1137, in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, token, storage_options, **config_kwargs)
+ 1135 if features is not None:
+ 1136 features = _fix_for_backward_compatible_features(features)
+-> 1137 dataset_module = dataset_module_factory(
+ 1138 path,
+ 1139 revision=revision,
+ 1140 download_config=download_config,
+ 1141 download_mode=download_mode,
+ 1142 data_dir=data_dir,
+ 1143 data_files=data_files,
+ 1144 cache_dir=cache_dir,
+ 1145 )
+ 1146 # Get dataset builder class
+ 1147 builder_kwargs = dataset_module.builder_kwargs
+
+File ~/.local/lib/python3.12/site-packages/datasets/load.py:1036, in dataset_module_factory(path, revision, download_config, download_mode, data_dir, data_files, cache_dir, **download_kwargs)
+ 1031 if isinstance(e1, FileNotFoundError):
+ 1032 raise FileNotFoundError(
+ 1033 f""Couldn't find any data file at {relative_to_absolute_path(path)}. ""
+ 1034 f""Couldn't find '{path}' on the Hugging Face Hub either: {type(e1).__name__}: {e1}""
+ 1035 ) from None
+-> 1036 raise e1 from None
+ 1037 else:
+ 1038 raise FileNotFoundError(f""Couldn't find any data file at {relative_to_absolute_path(path)}."")
+
+File ~/.local/lib/python3.12/site-packages/datasets/load.py:994, in dataset_module_factory(path, revision, download_config, download_mode, data_dir, data_files, cache_dir, **download_kwargs)
+ 986 try:
+ 987 api.hf_hub_download(
+ 988 repo_id=path,
+ 989 filename=filename,
+ (...)
+ 992 proxies=download_config.proxies,
+ 993 )
+--> 994 raise RuntimeError(f""Dataset scripts are no longer supported, but found {filename}"")
+ 995 except EntryNotFoundError:
+ 996 # Use the infos from the parquet export except in some cases:
+ 997 if data_dir or data_files or (revision and revision != ""main""):
+
+RuntimeError: Dataset scripts are no longer supported, but found conll2003.py
+Could someone tell me what is wrong?
","Try:
+from datasets import load_dataset
+dataset = load_dataset(""lhoestq/conll2003"")
+This is because support for trust_remote_code=True was removed in datasets library version 4.0.0 and later. You can work around this by using datasets that don’t rely on builder scripts (like the one shown above) or by downgrading the datasets library to version 3.6.0 or earlier.
I am trying to connect my custom domain, salsaqueen.club, to my Hugging Face Space. The status has been stuck in “pending” for several hours and the SSL certificate will not issue.
I have already done the following troubleshooting:
\nMy DNS is managed at GoDaddy.
\nThe www.mydomain.club subdomain is correctly configured as a CNAME record pointing to hf.space.
The root domain (mydomain.club) is correctly configured with a permanent 301 redirect to https://www.mydomain.club.
I have verified with public tools like DNSChecker.org that the CNAME record is propagating correctly worldwide.
\nI have already tried removing and re-adding the custom domain in the Hugging Face settings, but it remains stuck.
\nAll of my user-side configuration appears to be correct.
\nWhy is it not going live? Help is much appreciated
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-19T20:08:27.683Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 29, 'reads': 19, 'readers_count': 18, 'score': 153.6, 'yours': False, 'topic_id': 168554, 'topic_slug': 'custom-domain-stuck-on-pending', 'display_username': 'Jordan Glaus', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://DNSChecker.org', 'internal': False, 'reflection': False, 'title': 'DNS Checker - DNS Check Propagation Tool', 'clicks': 3}, {'url': 'https://discuss.huggingface.co/t/problem-custom-domain/168627/2', 'internal': True, 'reflection': True, 'title': 'Problem Custom domain', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104144, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/custom-domain-stuck-on-pending/168554/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242318, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-19T23:08:38.547Z', 'cooked': 'There seem to be several points that could potentially be improved.
\nCorrect setup for your case:
\nIn Hugging Face → Space → Settings → Custom domain
\nEnter www.salsaqueen.club (not the apex). The platform expects a subdomain CNAME pointed to hf.space. (Hugging Face)
In GoDaddy DNS (zone for salsaqueen.club)
\n\nApex behavior
\nUse GoDaddy’s HTTP 301 forwarding from salsaqueen.club → https://www.salsaqueen.club. Do not enable any forwarding on www. (GoDaddy)
Optional hardening
\nIf you later add CAA, include: CAA 0 issue ""letsencrypt.org"". Otherwise leave CAA absent. (Let’s Encrypt honors inherited or explicit CAA; conflicts can block issuance.) (Let’s Encrypt Community Support)
After you remove the www A records and leave only the single CNAME, delete and re-add the custom domain in Spaces. Status should move from Pending to Ready once validation sees the clean CNAME. (Hugging Face)
Hi @Jordamit thanks for reporting! We’re taking a look and I’ll update you soon.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-23T16:15:03.954Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 51.8, 'yours': False, 'topic_id': 168554, 'topic_slug': 'custom-domain-stuck-on-pending', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/custom-domain-stuck-on-pending/168554/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242445, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-23T19:34:12.074Z', 'cooked': 'Thank you! Megan.
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-23T19:34:12.074Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 31.8, 'yours': False, 'topic_id': 168554, 'topic_slug': 'custom-domain-stuck-on-pending', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/custom-domain-stuck-on-pending/168554/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242927, 'name': 'Jordan Glaus', 'username': 'Jordamit', 'avatar_template': '/user_avatar/discuss.huggingface.co/jordamit/{size}/54073_2.png', 'created_at': '2025-10-01T18:39:51.919Z', 'cooked': 'How this going? I’d love toast this up
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-10-01T18:39:51.919Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 168554, 'topic_slug': 'custom-domain-stuck-on-pending', 'display_username': 'Jordan Glaus', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 31941, 'username': 'meganariley', 'name': 'Megan Riley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104144, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/custom-domain-stuck-on-pending/168554/5', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243600, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-14T20:38:05.238Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-10-14T20:38:05.238Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168554, 'topic_slug': 'custom-domain-stuck-on-pending', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/custom-domain-stuck-on-pending/168554/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to connect my custom domain, salsaqueen.club, to my Hugging Face Space. The status has been stuck in “pending” for several hours and the SSL certificate will not issue.
I have already done the following troubleshooting:
+My DNS is managed at GoDaddy.
+The www.mydomain.club subdomain is correctly configured as a CNAME record pointing to hf.space.
The root domain (mydomain.club) is correctly configured with a permanent 301 redirect to https://www.mydomain.club.
I have verified with public tools like DNSChecker.org that the CNAME record is propagating correctly worldwide.
+I have already tried removing and re-adding the custom domain in the Hugging Face settings, but it remains stuck.
+All of my user-side configuration appears to be correct.
+Why is it not going live? Help is much appreciated
",Thank you! Megan.
+I don’t get it why Llama.cpp / GGML is so much faster than PyTorch,https://discuss.huggingface.co/t/i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch/168708,168708,9,2025-09-26 19:09:11.234000+00:00,"[{'id': 242642, 'name': 'Lorenzo Cesconetto', 'username': 'lorenzocc', 'avatar_template': '/user_avatar/discuss.huggingface.co/lorenzocc/{size}/54030_2.png', 'created_at': '2025-09-26T19:09:11.298Z', 'cooked': 'PyTorch offers a Python API, but the bulk of the processing is executed by the underlying C++ implementation (LibTorch).
\nGGML / Llama.cpp claims to be much faster because it was written in C/C++.
\nWhy is that the case? I don’t think the Python binding is adding too much overhead, shouldn’t they perform similarly?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-26T19:09:11.298Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 113, 'reads': 7, 'readers_count': 6, 'score': 396.4, 'yours': False, 'topic_id': 168708, 'topic_slug': 'i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch', 'display_username': 'Lorenzo Cesconetto', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104080, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch/168708/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242650, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-26T22:28:33.411Z', 'cooked': 'Rather than PyTorch being slow, I think the key to speed in Llama.cpp is likely its optimization of the generation strategy for CPU and GGUF quantized model weights. Hugging Face TGI, for example, uses PyTorch as one of its backend yet remains fast. Also, Python alone is slow and struggles with multi-core handling, but in scenarios where only the backend speed matters, it’s often not much of an issue.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-26T22:28:33.411Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 1.6, 'yours': False, 'topic_id': 168708, 'topic_slug': 'i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/text-generation-inference/conceptual/chunking', 'internal': False, 'reflection': False, 'title': 'TGI v3 overview', 'clicks': 4}, {'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/why_llamacpp_fast.md', 'internal': False, 'reflection': False, 'title': 'why_llamacpp_fast.md · John6666/forum1 at main', 'clicks': 4}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch/168708/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242670, 'name': 'Andrew Scott', 'username': 'Pimpcat-AU', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png', 'created_at': '2025-09-27T05:28:37.871Z', 'cooked': 'It is not about Python. It is about an inference only stack that is laser focused on CPU and cache behavior.
\nWhat llama dot cpp does that PyTorch usually does not on CPU
\nUses very aggressive quantization like four bit and five bit GGUF with per block scales and a layout that matches the matmul kernels. Fewer bytes moved is the main win on CPU.
\nShips hand tuned kernels that use SIMD like AVX2 or AVX512 on x86 and NEON on ARM with careful cache tiling and prefetch. These kernels are written for the model shapes that matter.
\nAvoids framework overhead. No autograd no shape polymorphism checks no dispatcher hops. Static shapes and static graph for inference.
\nMemory maps weights so cold start is faster and working sets stream in as needed. Very little extra copying.
\nThreads are pinned and scheduled for cache locality. The KV cache layout and rope math are optimized for batch size one and small batches.
\nFuses small ops so fewer passes over memory. Think dequantize and matmul in one sweep.
\nWhy PyTorch can look slower on CPU
\nIt is a general platform. The CPU path carries checks allocs layout conversions and dispatcher cost that help many models but cost cycles here.
\nIts quantized CPU kernels are improving but are not yet as specialized as llama dot cpp for this exact workload.
\nMany PyTorch setups keep weights in eight bit or sixteen bit and that alone moves two to four times more data through memory.
\nWhen PyTorch wins
\nOn GPU with cuBLAS and Tensor Cores a PyTorch model that uses half precision or better can outrun a CPU build by a large margin.
\nWith large batches or complex pipelines where the framework graph and kernels are already well optimized.
\nRule of thumb
\nOn CPU and small batch inference with strong quantization llama dot cpp usually wins. On GPU or with larger batches PyTorch often wins.
Reply generated by TD Ai.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-27T05:29:01.610Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 8, 'readers_count': 7, 'score': 56.6, 'yours': False, 'topic_id': 168708, 'topic_slug': 'i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch', 'display_username': 'Andrew Scott', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96276, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch/168708/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243466, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-12T20:00:45.129Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-12T20:00:45.129Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168708, 'topic_slug': 'i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/i-dont-get-it-why-llama-cpp-ggml-is-so-much-faster-than-pytorch/168708/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","PyTorch offers a Python API, but the bulk of the processing is executed by the underlying C++ implementation (LibTorch).
+GGML / Llama.cpp claims to be much faster because it was written in C/C++.
+Why is that the case? I don’t think the Python binding is adding too much overhead, shouldn’t they perform similarly?
","It is not about Python. It is about an inference only stack that is laser focused on CPU and cache behavior.
+What llama dot cpp does that PyTorch usually does not on CPU
+Uses very aggressive quantization like four bit and five bit GGUF with per block scales and a layout that matches the matmul kernels. Fewer bytes moved is the main win on CPU.
+Ships hand tuned kernels that use SIMD like AVX2 or AVX512 on x86 and NEON on ARM with careful cache tiling and prefetch. These kernels are written for the model shapes that matter.
+Avoids framework overhead. No autograd no shape polymorphism checks no dispatcher hops. Static shapes and static graph for inference.
+Memory maps weights so cold start is faster and working sets stream in as needed. Very little extra copying.
+Threads are pinned and scheduled for cache locality. The KV cache layout and rope math are optimized for batch size one and small batches.
+Fuses small ops so fewer passes over memory. Think dequantize and matmul in one sweep.
+Why PyTorch can look slower on CPU
+It is a general platform. The CPU path carries checks allocs layout conversions and dispatcher cost that help many models but cost cycles here.
+Its quantized CPU kernels are improving but are not yet as specialized as llama dot cpp for this exact workload.
+Many PyTorch setups keep weights in eight bit or sixteen bit and that alone moves two to four times more data through memory.
+When PyTorch wins
+On GPU with cuBLAS and Tensor Cores a PyTorch model that uses half precision or better can outrun a CPU build by a large margin.
+With large batches or complex pipelines where the framework graph and kernels are already well optimized.
+Rule of thumb
+On CPU and small batch inference with strong quantization llama dot cpp usually wins. On GPU or with larger batches PyTorch often wins.
Reply generated by TD Ai.
" +CUDA Deadlock while training DETR,https://discuss.huggingface.co/t/cuda-deadlock-while-training-detr/168917,168917,9,2025-10-05 11:29:15.125000+00:00,"[{'id': 243083, 'name': 'Ibrahim Dönmez', 'username': 'imetin', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/i/ecccb3/{size}.png', 'created_at': '2025-10-05T11:29:15.184Z', 'cooked': 'I was following the guideline for object detection in the guidelines to train DAB-DETR on my custom dataset. I have controlled collate_fn function and it worked as expected. On top of that, no issues with the dataset or the inputs format were spotted. The trainer and training arguments objects get initialized perfectly. However as the train method is called, I receive:
\n/usr/local/lib/python3.12/dist-packages/notebook/notebookapp.py:191: SyntaxWarning: invalid escape sequence \'\\/\'\n | |_| | \'_ \\/ _` / _` | _/ -_)\n\nafter this warning, nothing happens, no memory on gpu gets allocated. It just stays like that seeming to be running without doing anything. I am collab. When I try stopping the cell, it does not work and even restarting the runtime gets stuck, so only escape method is disconnecting from the runtime. Did anybody have similar experiences or know a solution?
\nTraining setting is as following:
\ntraining_args = TrainingArguments(\n output_dir=checkpoint_path_huggingface,\n num_train_epochs=30,\n fp16=False,\n per_device_train_batch_size=BATCH_SIZE,\n dataloader_num_workers=0,\n dataloader_pin_memory=False,\n disable_tqdm=False,\n report_to=None,\n learning_rate=1e-4,\n lr_scheduler_type=""cosine"",\n weight_decay=1e-4,\n max_grad_norm=0.1,\n metric_for_best_model=""eval_map"",\n greater_is_better=True,\n load_best_model_at_end=True,\n evaluation_strategy=""epoch"",\n save_strategy=""epoch"",\n save_total_limit=2,\n)\n\ntrainer = Trainer(\n model=model,\n args=training_args,\n train_dataset=train_dataset,\n eval_dataset=val_dataset,\n processing_class=processor,\n data_collator=collate_fn,\n compute_metrics=eval_compute_metrics_fn,\n)\nThat warning is the kind you can safely ignore. For example, if you’re storing your custom dataset on Google Drive, it seems to stall because it’s too slow.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-05T23:40:49.056Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 168917, 'topic_slug': 'cuda-deadlock-while-training-detr', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/colab_trainer_stall_without_message.md', 'internal': False, 'reflection': False, 'title': 'colab_trainer_stall_without_message.md · John6666/forum1 at main', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cuda-deadlock-while-training-detr/168917/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243454, 'name': 'Ibrahim Dönmez', 'username': 'imetin', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/i/ecccb3/{size}.png', 'created_at': '2025-10-11T22:35:30.260Z', 'cooked': 'Thank you very much, the issue got fixed.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-11T22:35:30.260Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 168917, 'topic_slug': 'cuda-deadlock-while-training-detr', 'display_username': 'Ibrahim Dönmez', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105041, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cuda-deadlock-while-training-detr/168917/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243455, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-11T22:35:30.344Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-11T22:35:30.344Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168917, 'topic_slug': 'cuda-deadlock-while-training-detr', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cuda-deadlock-while-training-detr/168917/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I was following the guideline for object detection in the guidelines to train DAB-DETR on my custom dataset. I have controlled collate_fn function and it worked as expected. On top of that, no issues with the dataset or the inputs format were spotted. The trainer and training arguments objects get initialized perfectly. However as the train method is called, I receive:
+/usr/local/lib/python3.12/dist-packages/notebook/notebookapp.py:191: SyntaxWarning: invalid escape sequence '\/'
+ | |_| | '_ \/ _` / _` | _/ -_)
+
+after this warning, nothing happens, no memory on gpu gets allocated. It just stays like that seeming to be running without doing anything. I am collab. When I try stopping the cell, it does not work and even restarting the runtime gets stuck, so only escape method is disconnecting from the runtime. Did anybody have similar experiences or know a solution?
+Training setting is as following:
+training_args = TrainingArguments(
+ output_dir=checkpoint_path_huggingface,
+ num_train_epochs=30,
+ fp16=False,
+ per_device_train_batch_size=BATCH_SIZE,
+ dataloader_num_workers=0,
+ dataloader_pin_memory=False,
+ disable_tqdm=False,
+ report_to=None,
+ learning_rate=1e-4,
+ lr_scheduler_type=""cosine"",
+ weight_decay=1e-4,
+ max_grad_norm=0.1,
+ metric_for_best_model=""eval_map"",
+ greater_is_better=True,
+ load_best_model_at_end=True,
+ evaluation_strategy=""epoch"",
+ save_strategy=""epoch"",
+ save_total_limit=2,
+)
+
+trainer = Trainer(
+ model=model,
+ args=training_args,
+ train_dataset=train_dataset,
+ eval_dataset=val_dataset,
+ processing_class=processor,
+ data_collator=collate_fn,
+ compute_metrics=eval_compute_metrics_fn,
+)
+That warning is the kind you can safely ignore. For example, if you’re storing your custom dataset on Google Drive, it seems to stall because it’s too slow.
" +WGET with Token not working,https://discuss.huggingface.co/t/wget-with-token-not-working/169024,169024,5,2025-10-08 09:03:54.478000+00:00,"[{'id': 243271, 'name': 'Lelièvre', 'username': 'RenanL', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/8dc957/{size}.png', 'created_at': '2025-10-08T09:03:54.532Z', 'cooked': 'Dear Hughingface Team,
\nI’m using runpod with the templates “ComfyUI - AI-Dock”.
\nIn JupyterLab I want to download a login protected model, the one from black-forest-labs/FLUX.1-Krea-dev.
\nwget used to work like that, I can download the model from my browser after login on my local pc.
\nwget --header=“Authorization: Bearer TOKEN” ``https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors
But I get
\n401 Unauthorized\nUsername/Password Authentication Failed.\nIf I add –debug at the end. I get:
\nDEBUG output created by Wget 1.21.2 on linux-gnu.\n\nReading HSTS entries from /home/user/.wget-hsts\nURI encoding = ‘UTF-8’\nConverted file name \'flux1-dev.safetensors\' (UTF-8) -> \'flux1-dev.safetensors\' (UTF-8)\n--2025-10-08 09:03:02-- https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors\nResolving huggingface.co (huggingface.co)... 52.84.217.103, 52.84.217.69, 52.84.217.102, ...\nCaching huggingface.co => 52.84.217.103 52.84.217.69 52.84.217.102 52.84.217.88 2600:9000:203d:6200:17:b174:6d00:93a1 2600:9000:203d:e000:17:b174:6d00:93a1 2600:9000:203d:8800:17:b174:6d00:93a1 2600:9000:203d:e800:17:b174:6d00:93a1 2600:9000:203d:9600:17:b174:6d00:93a1 2600:9000:203d:2400:17:b174:6d00:93a1 2600:9000:203d:ee00:17:b174:6d00:93a1 2600:9000:203d:6400:17:b174:6d00:93a1\nConnecting to huggingface.co (huggingface.co)|52.84.217.103|:443... connected.\nCreated socket 3.\nReleasing 0x000061bc69c86ec0 (new refcount 1).\nInitiating SSL handshake.\nHandshake successful; connected socket 3 to SSL handle 0x000061bc69c888a0\ncertificate:\n subject: CN=huggingface.co\n issuer: CN=Amazon RSA 2048 M02,O=Amazon,C=US\nX509 certificate successfully verified and matches host huggingface.co\n\n---request begin---\nGET /black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors HTTP/1.1\nHost: huggingface.co\nUser-Agent: Wget/1.21.2\nAccept: */*\nAccept-Encoding: identity\nConnection: Keep-Alive\nAuthorization: Bearer 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\n\n---request end---\nHTTP request sent, awaiting response... \n---response begin---\nHTTP/1.1 401 Unauthorized\nContent-Type: text/html; charset=utf-8\nContent-Length: 22349\nConnection: keep-alive\nDate: Wed, 08 Oct 2025 09:03:02 GMT\nETag: W/""574d-1eC4sA5Q/PbQ5YhsvC0L0NiNhEc""\nX-Powered-By: huggingface-moon\nRateLimit: ""pages"";r=999;t=66\nRateLimit-Policy: ""fixed window"";""pages"";q=1000;w=300\ncross-origin-opener-policy: same-origin\nReferrer-Policy: strict-origin-when-cross-origin\nX-Request-Id: Root=1-68e628c6-753c6a394bc274c7764e5a2f\nX-Error-Message: Invalid credentials in Authorization header\nx-frame-options: SAMEORIGIN\nX-Cache: Error from cloudfront\nVia: 1.1 fdd255cb127a7759980ee879db5de580.cloudfront.net (CloudFront)\nX-Amz-Cf-Pop: DFW59-P5\nX-Amz-Cf-Id: tZ4CtuVneK0RyHpWtL5_DbEc3eq4qqEMlGoXvt8V9CLxqmo2CX4puw==\n\n---response end---\n401 Unauthorized\nRegistered socket 3 for persistent reuse.\nDisabling further reuse of socket 3.\nClosed 3/SSL 0x000061bc69c888a0\n\nUsername/Password Authentication Failed.\nThank you for looking into that.
', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-08T09:03:54.532Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 15, 'reads': 6, 'readers_count': 5, 'score': 61.2, 'yours': False, 'topic_id': 169024, 'topic_slug': 'wget-with-token-not-working', 'display_username': 'Lelièvre', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105173, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/wget-with-token-not-working/169024/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243288, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-08T10:22:28.337Z', 'cooked': 'How about resolve instead of blob for now?
\nwget --header=""Authorization: Bearer TOKEN"" ""https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors""
resolve is solving the problem!
\nThank you so much for your help.
\nWhy I get blob instead of resolve in the url?
', 'post_number': 3, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-08T11:27:51.251Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 169024, 'topic_slug': 'wget-with-token-not-working', 'display_username': 'Lelièvre', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105173, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/wget-with-token-not-working/169024/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243299, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-08T11:38:28.728Z', 'cooked': 'blob is for web UI file-viewer URL. resolve is for file itself. Probably got mixed in from copy-pasting.
Need to check that!
\nThank you again.
', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-08T11:58:23.708Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 169024, 'topic_slug': 'wget-with-token-not-working', 'display_username': 'Lelièvre', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105173, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/wget-with-token-not-working/169024/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243326, 'name': 'Vu Hung Nguyen', 'username': 'vuhung', 'avatar_template': '/user_avatar/discuss.huggingface.co/vuhung/{size}/53965_2.png', 'created_at': '2025-10-08T22:23:11.995Z', 'cooked': 'In this context, is curl better than wget?
', 'post_number': 6, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-08T22:23:11.995Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 20.6, 'yours': False, 'topic_id': 169024, 'topic_slug': 'wget-with-token-not-working', 'display_username': 'Vu Hung Nguyen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103980, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/wget-with-token-not-working/169024/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243327, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-08T22:29:30.794Z', 'cooked': 'Yeah. Well, I think most people use curl. The HF sample also uses curl. Even in that case, though, you should probably use URLs with resolve in the default behavior.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-10-09T10:29:31.103Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 169024, 'topic_slug': 'wget-with-token-not-working', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/wget-with-token-not-working/169024/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Dear Hughingface Team,
+I’m using runpod with the templates “ComfyUI - AI-Dock”.
+In JupyterLab I want to download a login protected model, the one from black-forest-labs/FLUX.1-Krea-dev.
+wget used to work like that, I can download the model from my browser after login on my local pc.
+wget --header=“Authorization: Bearer TOKEN” ``https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors
But I get
+401 Unauthorized
+Username/Password Authentication Failed.
+If I add –debug at the end. I get:
+DEBUG output created by Wget 1.21.2 on linux-gnu.
+
+Reading HSTS entries from /home/user/.wget-hsts
+URI encoding = ‘UTF-8’
+Converted file name 'flux1-dev.safetensors' (UTF-8) -> 'flux1-dev.safetensors' (UTF-8)
+--2025-10-08 09:03:02-- https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors
+Resolving huggingface.co (huggingface.co)... 52.84.217.103, 52.84.217.69, 52.84.217.102, ...
+Caching huggingface.co => 52.84.217.103 52.84.217.69 52.84.217.102 52.84.217.88 2600:9000:203d:6200:17:b174:6d00:93a1 2600:9000:203d:e000:17:b174:6d00:93a1 2600:9000:203d:8800:17:b174:6d00:93a1 2600:9000:203d:e800:17:b174:6d00:93a1 2600:9000:203d:9600:17:b174:6d00:93a1 2600:9000:203d:2400:17:b174:6d00:93a1 2600:9000:203d:ee00:17:b174:6d00:93a1 2600:9000:203d:6400:17:b174:6d00:93a1
+Connecting to huggingface.co (huggingface.co)|52.84.217.103|:443... connected.
+Created socket 3.
+Releasing 0x000061bc69c86ec0 (new refcount 1).
+Initiating SSL handshake.
+Handshake successful; connected socket 3 to SSL handle 0x000061bc69c888a0
+certificate:
+ subject: CN=huggingface.co
+ issuer: CN=Amazon RSA 2048 M02,O=Amazon,C=US
+X509 certificate successfully verified and matches host huggingface.co
+
+---request begin---
+GET /black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors HTTP/1.1
+Host: huggingface.co
+User-Agent: Wget/1.21.2
+Accept: */*
+Accept-Encoding: identity
+Connection: Keep-Alive
+Authorization: Bearer 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
+
+---request end---
+HTTP request sent, awaiting response...
+---response begin---
+HTTP/1.1 401 Unauthorized
+Content-Type: text/html; charset=utf-8
+Content-Length: 22349
+Connection: keep-alive
+Date: Wed, 08 Oct 2025 09:03:02 GMT
+ETag: W/""574d-1eC4sA5Q/PbQ5YhsvC0L0NiNhEc""
+X-Powered-By: huggingface-moon
+RateLimit: ""pages"";r=999;t=66
+RateLimit-Policy: ""fixed window"";""pages"";q=1000;w=300
+cross-origin-opener-policy: same-origin
+Referrer-Policy: strict-origin-when-cross-origin
+X-Request-Id: Root=1-68e628c6-753c6a394bc274c7764e5a2f
+X-Error-Message: Invalid credentials in Authorization header
+x-frame-options: SAMEORIGIN
+X-Cache: Error from cloudfront
+Via: 1.1 fdd255cb127a7759980ee879db5de580.cloudfront.net (CloudFront)
+X-Amz-Cf-Pop: DFW59-P5
+X-Amz-Cf-Id: tZ4CtuVneK0RyHpWtL5_DbEc3eq4qqEMlGoXvt8V9CLxqmo2CX4puw==
+
+---response end---
+401 Unauthorized
+Registered socket 3 for persistent reuse.
+Disabling further reuse of socket 3.
+Closed 3/SSL 0x000061bc69c888a0
+
+Username/Password Authentication Failed.
+Thank you for looking into that.
","How about resolve instead of blob for now?
+wget --header=""Authorization: Bearer TOKEN"" ""https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors""
Hi everyone,
\nI’m developing a pronunciation app for deaf users learning Korean on iOS (Swift) and need to capture actual phonetic pronunciation as text.
\nIn Korean, the written form differs from the actual pronunciation due to phonological rules.
\nExample:
\nAnother example:
\nAll STT systems output standard orthography, not phonetic transcription. For deaf users learning pronunciation, they need to see exactly how words sound (e.g., “모교일”), not the standard spelling (“목요일”).
\nconfidence scores but not phonetic outputIs it possible to get phonetic transcription (not standard orthography) from speech on iOS?
\nCan Wav2Vec2 or similar models output phonetic text instead of standard spelling? Can this be converted to Core ML?
\nAre there Korean-specific ASR models trained to output phonetic transcription rather than standard orthography?
\nHybrid approach? Could I combine:
\nStandard STT (Apple Speech) → “목요일”
\nThis text will be hidden
\nText-to-phonetic converter (g2pK) → “모교일”
\nBut how to handle actual mispronunciations?
\nIs this fundamentally impossible? Do all modern ASR systems inherently output standard orthography?
\nThis is for accessibility. Deaf users learning Korean need to understand that “목요일” is pronounced “모교일”, not “목-요-일” (syllable by syllable).
\nStandard STT’s conversion to orthography is exactly what I need to avoid.
\nIf phonetic transcription from speech is impossible, what are realistic alternatives for teaching pronunciation to deaf users?
\nThank you for any insights!
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-08T05:45:07.760Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 169014, 'topic_slug': 'how-to-extract-actual-phonetic-pronunciation-as-text-on-ios-korean-phonetic-transcription', 'display_username': 'Moon Ho', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105210, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-extract-actual-phonetic-pronunciation-as-text-on-ios-korean-phonetic-transcription/169014/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243264, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-08T08:23:00.431Z', 'cooked': 'I don’t know Swift very well, so I’ll just put the resources here for now…
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-08T08:23:00.431Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 169014, 'topic_slug': 'how-to-extract-actual-phonetic-pronunciation-as-text-on-ios-korean-phonetic-transcription', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/ios_phonetic_transcription.md', 'internal': False, 'reflection': False, 'title': 'ios_phonetic_transcription.md · John6666/forum1 at main', 'clicks': 3}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-extract-actual-phonetic-pronunciation-as-text-on-ios-korean-phonetic-transcription/169014/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243307, 'name': 'Moon Ho', 'username': 'moonshiro', 'avatar_template': '/user_avatar/discuss.huggingface.co/moonshiro/{size}/54632_2.png', 'created_at': '2025-10-08T13:10:27.894Z', 'cooked': 'Thank you. It really helped me a lot.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-09T01:11:02.459Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 169014, 'topic_slug': 'how-to-extract-actual-phonetic-pronunciation-as-text-on-ios-korean-phonetic-transcription', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-extract-actual-phonetic-pronunciation-as-text-on-ios-korean-phonetic-transcription/169014/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I’m developing a pronunciation app for deaf users learning Korean on iOS (Swift) and need to capture actual phonetic pronunciation as text.
+In Korean, the written form differs from the actual pronunciation due to phonological rules.
+Example:
+Another example:
+All STT systems output standard orthography, not phonetic transcription. For deaf users learning pronunciation, they need to see exactly how words sound (e.g., “모교일”), not the standard spelling (“목요일”).
+confidence scores but not phonetic outputIs it possible to get phonetic transcription (not standard orthography) from speech on iOS?
+Can Wav2Vec2 or similar models output phonetic text instead of standard spelling? Can this be converted to Core ML?
+Are there Korean-specific ASR models trained to output phonetic transcription rather than standard orthography?
+Hybrid approach? Could I combine:
+Standard STT (Apple Speech) → “목요일”
+This text will be hidden
+Text-to-phonetic converter (g2pK) → “모교일”
+But how to handle actual mispronunciations?
+Is this fundamentally impossible? Do all modern ASR systems inherently output standard orthography?
+This is for accessibility. Deaf users learning Korean need to understand that “목요일” is pronounced “모교일”, not “목-요-일” (syllable by syllable).
+Standard STT’s conversion to orthography is exactly what I need to avoid.
+If phonetic transcription from speech is impossible, what are realistic alternatives for teaching pronunciation to deaf users?
+Thank you for any insights!
","I don’t know Swift very well, so I’ll just put the resources here for now…
" +NonMatchingSplitsSizesError,https://discuss.huggingface.co/t/nonmatchingsplitssizeserror/30033,30033,10,2023-01-19 20:12:35.014000+00:00,"[{'id': 55242, 'name': 'Sundeep', 'username': 'sl02', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/ba9def/{size}.png', 'created_at': '2023-01-19T20:12:35.084Z', 'cooked': 'I created a custom script which splits the raw file into train/test split on the fly. The script works with the default arguments. However, when I change the test_size ratio which I pass via load_dataset(), it fails with the following error
Traceback (most recent call last): \n File ""<stdin>"", line 1, in <module>\n File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/load.py"", line 1757, in load_dataset\n builder_instance.download_and_prepare(\n File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/builder.py"", line 860, in download_and_prepare\n self._download_and_prepare(\n File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/builder.py"", line 1611, in _download_and_prepare\n super()._download_and_prepare(\n File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/builder.py"", line 971, in _download_and_prepare\n verify_splits(self.info.splits, split_dict)\n File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/utils/info_utils.py"", line 74, in verify_splits\n raise NonMatchingSplitsSizesError(str(bad_splits))\ndatasets.utils.info_utils.NonMatchingSplitsSizesError\nIt fails the integrity check as expected. The Build and load doesn’t show how to update the checks. I thought, using the download_mode=force_redownload argument in load_dataset() would fix it but it throws the same error as shown above. How do I resolve this?
Hi @sl02 ! Is test_size a custom builder parameter you define in your loading script?
You can set ignore_verifications=True param in load_dataset to skip splits sizes verification.
Also note that Dataset object has .train_test_split() method, probably it might be useful for your case.
Hi @polinaeterna
\nYes. test_size is a parameter. Sure with the ignore_verifications=True parameter it works. But I would like to know how, for other datasets when it changes at the source, do you update the information; The instructions in the document, to which I provide a link in the above thread, doesn’t explain this clearly.
I am doing a group shuffle split because I have to ensure no overlap in the id column in the respective splits.
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2023-01-27T13:14:44.170Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 1, 'incoming_link_count': 85, 'reads': 148, 'readers_count': 147, 'score': 459.6, 'yours': False, 'topic_id': 30033, 'topic_slug': 'nonmatchingsplitssizeserror', 'display_username': 'Sundeep', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 8429, 'username': 'polinaeterna', 'name': 'Polina Kazakova', 'avatar_template': '/user_avatar/discuss.huggingface.co/polinaeterna/{size}/19055_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 12315, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/nonmatchingsplitssizeserror/30033/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 56173, 'name': 'Polina Kazakova', 'username': 'polinaeterna', 'avatar_template': '/user_avatar/discuss.huggingface.co/polinaeterna/{size}/19055_2.png', 'created_at': '2023-01-27T17:56:14.846Z', 'cooked': '@sl02
\nWhen you load your dataset locally for the first time, it creates dataset_info.json file under its cache folder, the file contains all these splits info (like num_examples, num_bytes, etc.). If you regenerate the dataset while the script is unchanged (for example, run load_dataset with download_mode=""reuse_cache_if_exists""), it performs verifications against this file.
We used to have dataset_info.json files in datasets repositories on the Hub (so, not just in a local cache folder) to verify splits info on the first download but now it’s deprecated, we use README.md instead for storing these numbers.
\nTo (re)compute these numbers automatically and dump them to a README.md file, one should run datasets-cli test your_dataset --save_info. And as it’s done manually, it depends on datasets’ authors if they update and push this info or not as it’s not required.
\nHope it’s more or less clear, feel free to ask any questions if it’s not
@polinaeterna
\nThanks for clearing that up!
Note that you could get this error when you try and download an updated dataset without using the cache. E.g.,
\ndataset = load_dataset(url, download_mode=“force_redownload”)
If the underlying dataset has been updated there can be a miss-match between the number of read records and what is read from the cache. You can read about the cache here, Cache management.
', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2023-09-13T19:07:17.850Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 26, 'reads': 85, 'readers_count': 84, 'score': 147.0, 'yours': False, 'topic_id': 30033, 'topic_slug': 'nonmatchingsplitssizeserror', 'display_username': 'Adam Hjerpe', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/datasets/cache', 'internal': False, 'reflection': False, 'title': 'Cache management', 'clicks': 123}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 27951, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/nonmatchingsplitssizeserror/30033/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243312, 'name': 'Albert Zeyer', 'username': 'albertzeyer', 'avatar_template': '/user_avatar/discuss.huggingface.co/albertzeyer/{size}/46906_2.png', 'created_at': '2025-10-08T16:51:31.810Z', 'cooked': '\nThis does not work anymore. I think now you have to use verification_mode=VerificationMode.NO_CHECKS.
I created a custom script which splits the raw file into train/test split on the fly. The script works with the default arguments. However, when I change the test_size ratio which I pass via load_dataset(), it fails with the following error
Traceback (most recent call last):
+ File ""<stdin>"", line 1, in <module>
+ File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/load.py"", line 1757, in load_dataset
+ builder_instance.download_and_prepare(
+ File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/builder.py"", line 860, in download_and_prepare
+ self._download_and_prepare(
+ File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/builder.py"", line 1611, in _download_and_prepare
+ super()._download_and_prepare(
+ File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/builder.py"", line 971, in _download_and_prepare
+ verify_splits(self.info.splits, split_dict)
+ File ""/Users/home/.local/share/virtualenvs/1717-yQ3Y_lVD/lib/python3.8/site-packages/datasets/utils/info_utils.py"", line 74, in verify_splits
+ raise NonMatchingSplitsSizesError(str(bad_splits))
+datasets.utils.info_utils.NonMatchingSplitsSizesError
+It fails the integrity check as expected. The Build and load doesn’t show how to update the checks. I thought, using the download_mode=force_redownload argument in load_dataset() would fix it but it throws the same error as shown above. How do I resolve this?
@sl02
+When you load your dataset locally for the first time, it creates dataset_info.json file under its cache folder, the file contains all these splits info (like num_examples, num_bytes, etc.). If you regenerate the dataset while the script is unchanged (for example, run load_dataset with download_mode=""reuse_cache_if_exists""), it performs verifications against this file.
We used to have dataset_info.json files in datasets repositories on the Hub (so, not just in a local cache folder) to verify splits info on the first download but now it’s deprecated, we use README.md instead for storing these numbers.
+To (re)compute these numbers automatically and dump them to a README.md file, one should run datasets-cli test your_dataset --save_info. And as it’s done manually, it depends on datasets’ authors if they update and push this info or not as it’s not required.
+Hope it’s more or less clear, feel free to ask any questions if it’s not
As we dive deeper into 2025 with more complex AI workflows, testing APIs for model deployments has become crucial. If you’re tired of Postman’s syncing issues or bloated interface when working with Hugging Face endpoints, you’re not alone. I’ve been exploring the best Postman alternatives optimized for AI devs like us focusing on speed, offline capabilities, and seamless integration with tools like Transformers library.
\nHere’s my quick rundown of top picks:
\nBruno: Lightweight and Git-friendly, perfect for version-controlling your API requests during model fine-tuning sessions. Great for solo AI experimenters.
\nHoppscotch: Open-source and browser-based—ideal for quick tests on Hugging Face Spaces without installing anything.
\nInsomnia: Robust for GraphQL and REST APIs, with strong support for environment variables that shine in multi-model testing.
\nThunder Client: VS Code extension that’s a game-changer if you’re scripting API calls alongside your Python notebooks.
\nBut after testing them all with real Hugging Face inference endpoints, Apidog emerges as my number one go-to. Its all-in-one platform handles API design, mocking, and debugging with AI-specific features like auto-generated OpenAPI docs tailored for ML pipelines saving me hours on collaborative projects. Plus, it’s fully offline-capable, so no more cloud dependency during sensitive model evals.
\nWhat are you using for Postman alternatives in your AI API workflows? Share below—let’s crowdsource the ultimate stack for 2025!
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-10-07T04:51:20.660Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 13, 'reads': 5, 'readers_count': 4, 'score': 71.0, 'yours': False, 'topic_id': 168983, 'topic_slug': 'best-postman-alternatives-for-ai-api-testing-in-2025', 'display_username': 'luc dev', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99922, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/best-postman-alternatives-for-ai-api-testing-in-2025/168983/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243203, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-07T08:23:41.942Z', 'cooked': 'For now I just gathered resources…
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-10-07T08:29:08.047Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 65.8, 'yours': False, 'topic_id': 168983, 'topic_slug': 'best-postman-alternatives-for-ai-api-testing-in-2025', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/postman_alternative_1.md', 'internal': False, 'reflection': False, 'title': 'postman_alternative_1.md · John6666/forum1 at main', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/best-postman-alternatives-for-ai-api-testing-in-2025/168983/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243257, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-08T07:40:22.307Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-10-08T07:40:22.307Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168983, 'topic_slug': 'best-postman-alternatives-for-ai-api-testing-in-2025', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/best-postman-alternatives-for-ai-api-testing-in-2025/168983/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","As we dive deeper into 2025 with more complex AI workflows, testing APIs for model deployments has become crucial. If you’re tired of Postman’s syncing issues or bloated interface when working with Hugging Face endpoints, you’re not alone. I’ve been exploring the best Postman alternatives optimized for AI devs like us focusing on speed, offline capabilities, and seamless integration with tools like Transformers library.
+Here’s my quick rundown of top picks:
+Bruno: Lightweight and Git-friendly, perfect for version-controlling your API requests during model fine-tuning sessions. Great for solo AI experimenters.
+Hoppscotch: Open-source and browser-based—ideal for quick tests on Hugging Face Spaces without installing anything.
+Insomnia: Robust for GraphQL and REST APIs, with strong support for environment variables that shine in multi-model testing.
+Thunder Client: VS Code extension that’s a game-changer if you’re scripting API calls alongside your Python notebooks.
+But after testing them all with real Hugging Face inference endpoints, Apidog emerges as my number one go-to. Its all-in-one platform handles API design, mocking, and debugging with AI-specific features like auto-generated OpenAPI docs tailored for ML pipelines saving me hours on collaborative projects. Plus, it’s fully offline-capable, so no more cloud dependency during sensitive model evals.
+What are you using for Postman alternatives in your AI API workflows? Share below—let’s crowdsource the ultimate stack for 2025!
","For now I just gathered resources…
" +Smolagents with Azure AI Foundry OpenAI model and DefaultAzureCredential or ManagedIdentity,https://discuss.huggingface.co/t/smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity/168997,168997,13,2025-10-07 11:54:02.248000+00:00,"[{'id': 243213, 'name': 'Ingo Villnow', 'username': 'IngoTB303', 'avatar_template': '/user_avatar/discuss.huggingface.co/ingotb303/{size}/28183_2.png', 'created_at': '2025-10-07T11:54:02.327Z', 'cooked': 'Hi there,
\ncurrently I use smolagents with AzureOpenAIServerModel() and an API key. Now I have to switch to Active Directory authentication with DefaultAzureCredential or ManagedIdentityCredential, but with smolagent’s AzureOpenAIServerModel or OpenAIServerModel it is not working. Any idea on that? I would like to keep smolagents as framework for my agents.
\nmodel = AzureOpenAIServerModel(\n model_id = AZURE_OPENAI_MODEL,\n azure_endpoint = AZURE_OPENAI_ENDPOINT,\n api_key = AZURE_OPENAI_API_KEY,\n api_version = OPENAI_API_VERSION \n)\nThanks and BR,
\nIngo
There seem to be multiple possible causes.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-07T12:41:49.132Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 168997, 'topic_slug': 'smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/smolagents_azure_not_work.md', 'internal': False, 'reflection': False, 'title': 'smolagents_azure_not_work.md · John6666/forum1 at main', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity/168997/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243224, 'name': 'Ingo Villnow', 'username': 'IngoTB303', 'avatar_template': '/user_avatar/discuss.huggingface.co/ingotb303/{size}/28183_2.png', 'created_at': '2025-10-07T14:28:01.792Z', 'cooked': 'Hi, I found out, how it works: forward the needed parameter as client_kwargs:
\nfrom azure.identity import DefaultAzureCredential, get_bearer_token_provider\n...\n\nclient_kwargs = {}\nif auth_mode == ""aad"": \n scope = os.getenv(""AZURE_OPENAI_SCOPE"", ""https://cognitiveservices.azure.com/.default"")\n credential = DefaultAzureCredential()\n client_kwargs[""azure_ad_token_provider""] = get_bearer_token_provider(credential, scope)\nelse: \n # default back to API key authentication\n api_key = os.getenv(""AZURE_OPENAI_API_KEY"")\nBest regards,
\nIngo
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-07T14:28:01.792Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168997, 'topic_slug': 'smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity', 'display_username': 'Ingo Villnow', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 46776, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity/168997/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243244, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-08T02:28:22.251Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-08T02:28:22.251Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168997, 'topic_slug': 'smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/smolagents-with-azure-ai-foundry-openai-model-and-defaultazurecredential-or-managedidentity/168997/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi there,
+currently I use smolagents with AzureOpenAIServerModel() and an API key. Now I have to switch to Active Directory authentication with DefaultAzureCredential or ManagedIdentityCredential, but with smolagent’s AzureOpenAIServerModel or OpenAIServerModel it is not working. Any idea on that? I would like to keep smolagents as framework for my agents.
+model = AzureOpenAIServerModel(
+ model_id = AZURE_OPENAI_MODEL,
+ azure_endpoint = AZURE_OPENAI_ENDPOINT,
+ api_key = AZURE_OPENAI_API_KEY,
+ api_version = OPENAI_API_VERSION
+)
+Thanks and BR,
+Ingo
There seem to be multiple possible causes.
" +Storage Quota Out of limit,https://discuss.huggingface.co/t/storage-quota-out-of-limit/168966,168966,5,2025-10-06 14:01:05.839000+00:00,"[{'id': 243169, 'name': 'Amaal Anoos', 'username': 'amaalanoosucs', 'avatar_template': '/user_avatar/discuss.huggingface.co/amaalanoosucs/{size}/54178_2.png', 'created_at': '2025-10-06T14:01:05.907Z', 'cooked': 'Hi Guys,
\nI’m on the free plan, and I have an issue with my storage limit. My current usage is showing as 35.6 GB/-146.14 GB. I never subscribed to the PRO as well. So why am I having -146.14 GB?
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-06T14:01:05.907Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 8, 'readers_count': 7, 'score': 26.6, 'yours': False, 'topic_id': 168966, 'topic_slug': 'storage-quota-out-of-limit', 'display_username': 'Amaal Anoos', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104321, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/storage-quota-out-of-limit/168966/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243171, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-06T14:53:07.276Z', 'cooked': 'here. Organization storage limit is negative 3 TB
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-06T14:53:07.276Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 1.6, 'yours': False, 'topic_id': 168966, 'topic_slug': 'storage-quota-out-of-limit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/organization-storage-limit-is-negative-3-tb/168909', 'internal': True, 'reflection': False, 'title': 'Organization storage limit is negative 3 TB', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/storage-quota-out-of-limit/168966/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243191, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-07T02:35:15.926Z', 'cooked': 'Today, I confirmed the fix in my environment. I think it’s probably fixed for others too…
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-07T02:35:15.926Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.4, 'yours': False, 'topic_id': 168966, 'topic_slug': 'storage-quota-out-of-limit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/storage-quota-out-of-limit/168966/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243210, 'name': 'Amaal Anoos', 'username': 'amaalanoosucs', 'avatar_template': '/user_avatar/discuss.huggingface.co/amaalanoosucs/{size}/54178_2.png', 'created_at': '2025-10-07T10:12:13.181Z', 'cooked': 'Hey John,
\nYes, the issue has been resolved. Thanks, for the heads up
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-07T10:12:13.181Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 168966, 'topic_slug': 'storage-quota-out-of-limit', 'display_username': 'Amaal Anoos', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104321, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/storage-quota-out-of-limit/168966/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243242, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-07T22:12:28.896Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-10-07T22:12:28.896Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168966, 'topic_slug': 'storage-quota-out-of-limit', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/storage-quota-out-of-limit/168966/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi Guys,
+I’m on the free plan, and I have an issue with my storage limit. My current usage is showing as 35.6 GB/-146.14 GB. I never subscribed to the PRO as well. So why am I having -146.14 GB?
+","Today, I confirmed the fix in my environment. I think it’s probably fixed for others too…
" +Error 404 when downloading the tokenizer,https://discuss.huggingface.co/t/error-404-when-downloading-the-tokenizer/168993,168993,9,2025-10-07 08:40:03.319000+00:00,"[{'id': 243207, 'name': 'Stefano', 'username': 'stefra', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/a9a28c/{size}.png', 'created_at': '2025-10-07T08:40:03.383Z', 'cooked': 'When I try to execute the following lines of code:
\nquantization_config = BitsAndBytesConfig(load_in_8bit=True)
\ntokenizer = AutoTokenizer.from_pretrained(model_id)
\nmodel = AutoModelForCausalLM.from_pretrained(
\nmodel_id,
\ndevice_map=“auto”,
\nquantization_config=quantization_config
\n)
The tokenizer raises a 404 Client Error: Not Found, specifically:
\n“Entry Not Found for URL: https://huggingface.co/api/models/Qwen/Qwen2.5-7B-Instruct/tree/main/additional_chat_templates?recursive=false&expand=false.
\nadditional_chat_templates does not exist on ‘main’.”
The libraries I am using are:
\ntokenizers == 0.21.2
transformers == 4.53.3
bitsandbytes == 0.48.1
Is there anything I can do to fix this issue? Could it be related to a version mismatch? Any advice would be appreciated.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-10-07T08:40:03.383Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 595, 'reads': 12, 'readers_count': 11, 'score': 2142.0, 'yours': False, 'topic_id': 168993, 'topic_slug': 'error-404-when-downloading-the-tokenizer', 'display_username': 'Stefano', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/api/models/Qwen/Qwen2.5-7B-Instruct/tree/main/additional_chat_templates?recursive=false&expand=false', 'internal': False, 'reflection': False, 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 105159, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-404-when-downloading-the-tokenizer/168993/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243209, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-07T09:34:58.688Z', 'cooked': 'Seems a resolved bug of Transformers. Try upgrade pip install -U transformers
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-10-07T21:35:22.053Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 10, 'readers_count': 9, 'score': 16.6, 'yours': False, 'topic_id': 168993, 'topic_slug': 'error-404-when-downloading-the-tokenizer', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/error-404-when-downloading-the-tokenizer/168993/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","When I try to execute the following lines of code:
+quantization_config = BitsAndBytesConfig(load_in_8bit=True)
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+model_id,
+device_map=“auto”,
+quantization_config=quantization_config
+)
The tokenizer raises a 404 Client Error: Not Found, specifically:
+“Entry Not Found for URL: https://huggingface.co/api/models/Qwen/Qwen2.5-7B-Instruct/tree/main/additional_chat_templates?recursive=false&expand=false.
+additional_chat_templates does not exist on ‘main’.”
The libraries I am using are:
+tokenizers == 0.21.2
transformers == 4.53.3
bitsandbytes == 0.48.1
Is there anything I can do to fix this issue? Could it be related to a version mismatch? Any advice would be appreciated.
","Seems a resolved bug of Transformers. Try upgrade pip install -U transformers
Hi there,
\nI’m new both on this forum and huggingface world. Please go easy on me
\nI have a question to ask. I want to use auto train for fine tune a model like meta-llama/Llama-3.1-8B-Instruct. I have a data set which is in alpaca model with instruction, input and output columns.
Questions are;
\nI couldn’t find a good document or example in order to learn how to fine tune a model with using this type of model.
\nNone of the information buttons are working on the Auto Train screen like the one above task or parameter combo-box.
\nHow can I put more fields in column mapping section? There is only one right now. I think I should put instruction, input and output columns.
\nIf there is any good documentation, please share it with me. So, I can started to learn some stuff.
\n\nBest regards,
\nYunus Emre
Hmm… Try this. And for AutoTrain CSV data format.
', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-09-26T23:14:08.034Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 11.6, 'yours': False, 'topic_id': 168711, 'topic_slug': 'auto-train-with-alpaca-model-data-set', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/autotrain/en/tasks/llm_finetuning', 'internal': False, 'reflection': False, 'title': 'LLM Finetuning with AutoTrain Advanced', 'clicks': 3}, {'url': 'https://discuss.huggingface.co/t/autotrain-csv-data-format/63305', 'internal': True, 'reflection': False, 'title': 'AutoTrain csv data format', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/auto-train-with-alpaca-model-data-set/168711/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242920, 'name': 'Yunus Emre BAYRAM', 'username': 'ynsbyrm', 'avatar_template': '/user_avatar/discuss.huggingface.co/ynsbyrm/{size}/54307_2.png', 'created_at': '2025-10-01T17:59:16.814Z', 'cooked': 'Hi @John6666 ,
\nThank you for your response. I’ve made some tries regarding the links which you’ve shared. I believe it is better now but I still have some questions. If you redirect me it would be really helpful.
\nFor LLM SFT task I need to combine the columns from data set and put them in one column as text in the csv. The point which I don’t understand how LLM will understand which column means what? I saw there are few other data sets in here for example one of them has 3 columns but other one has 7. Is there anyway to differentiate which data set should use in which case or is this requires a knowledge of data scientists?
\nBest regards,
\nYunus
I don’t have any data science knowledge whatsoever, but I think we can manage if we just do some basic preprocessing in Python… Functions for data processing and shaping are usually available somewhere in the libraries.
\nUse one rendered text column for SFT. Do not map instruction/input/output separately. Convert your rows to the model’s chat format, save as a single-column dataset, and map text → text in AutoTrain. (Hugging Face)
meta-llama/Llama-3.1-8B-Instruct. (Hugging Face)SFTTrainer accepts either:
\n{""text"": ""...final rendered conversation...""}, or{""prompt"": ""..."", ""completion"": ""...""}.text column for chat SFT. (Hugging Face)instruction + (""\\n\\n"" + input if present); assistant = output.apply_chat_template(messages, tokenize=False, add_generation_prompt=False).from datasets import load_dataset\nfrom transformers import AutoTokenizer\nimport pandas as pd\n\ntok = AutoTokenizer.from_pretrained(""meta-llama/Llama-3.1-8B-Instruct"")\n\ndef render_row(r):\n user = r[""instruction""] + ((""\\n\\n"" + r[""input""]) if r.get(""input"") else """")\n messages = [{""role"":""user"",""content"":user},\n {""role"":""assistant"",""content"":r[""output""]}]\n return tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=False)\n\nds = load_dataset(""tatsu-lab/alpaca"", split=""train"") # replace with your data\ndf = pd.DataFrame({""text"": [render_row(x) for x in ds]})\ndf.to_csv(""autotrain_llm_sft.csv"", index=False)\napply_chat_template ensures the exact prompt tokens and headers the model expects. (Hugging Face)
UI: upload CSV/JSONL, set Column Mapping → text → text, choose LLM finetuning → SFT. (Hugging Face)
\nCLI (reliable, explicit):
pip install autotrain-advanced\n\nautotrain llm \\\n --train \\\n --project-name llama31-alpaca-sft \\\n --model meta-llama/Llama-3.1-8B-Instruct \\\n --data-path ./ \\\n --train-split train \\\n --text-column text \\\n --trainer sft \\\n --use-peft \\\n --lora-r 16 --lora-alpha 32 --lora-dropout 0.05 \\\n --batch-size 4 --gradient-accumulation 8 \\\n --lr 2e-4 --epochs 3 --bf16 \\\n --max-seq-length 4096\nFlags mirror documented AutoTrain usage. Adjust batch and GA for VRAM. (Hugging Face)
\nAt generation, build messages and call the same tokenizer’s chat template to format the prompt before generate. Template mismatches degrade outputs. Llama 3.1 has known header nuances; verify your output. (Hugging Face)
Only if you pick a different trainer or format:
\nprompt and completion. (Hugging Face)prompt, chosen, rejected. AutoTrain exposes those roles in column mapping. (Hugging Face)--text-column text. (Hugging Face)AutoTrain LLM finetuning and column mapping, TRL SFT dataset formats, and chat templating docs. (Hugging Face)
', 'post_number': 4, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-01T21:59:45.363Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 1.2, 'yours': False, 'topic_id': 168711, 'topic_slug': 'auto-train-with-alpaca-model-data-set', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://www.llama.com/docs/model-cards-and-prompt-formats/llama3_1/', 'internal': False, 'reflection': False, 'clicks': 0}, {'url': 'https://huggingface.co/docs/autotrain/en/col_map', 'internal': False, 'reflection': False, 'title': 'Understanding Column Mapping', 'clicks': 0}, {'url': 'https://huggingface.co/docs/trl/en/sft_trainer', 'internal': False, 'reflection': False, 'title': 'SFT Trainer', 'clicks': 0}, {'url': 'https://huggingface.co/docs/autotrain/en/tasks/llm_finetuning', 'internal': False, 'reflection': False, 'title': 'LLM Finetuning with AutoTrain Advanced', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/en/chat_templating', 'internal': False, 'reflection': False, 'title': 'Chat templates', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/v4.51.1/chat_templating', 'internal': False, 'reflection': False, 'title': 'Templates', 'clicks': 0}, {'url': 'https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct', 'internal': False, 'reflection': False, 'title': 'meta-llama/Llama-3.1-8B-Instruct · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/auto-train-with-alpaca-model-data-set/168711/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242936, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-01T23:07:44.757Z', 'cooked': 'For SFT and its practical implementation, the Smol course provides a concise overview of the entire process, so I recommend giving it a quick read.
', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-10-01T23:07:44.757Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 168711, 'topic_slug': 'auto-train-with-alpaca-model-data-set', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/learn/smol-course/unit0/1', 'internal': False, 'reflection': False, 'title': 'Welcome to the 🤗 smol-course - Hugging Face a smol course', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/auto-train-with-alpaca-model-data-set/168711/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243019, 'name': 'Yunus Emre BAYRAM', 'username': 'ynsbyrm', 'avatar_template': '/user_avatar/discuss.huggingface.co/ynsbyrm/{size}/54307_2.png', 'created_at': '2025-10-03T08:31:23.922Z', 'cooked': 'Hi @John6666 ,
\nGreat explanation and these are wonderful links. I’m feel like enlightened. Even I started to following that smol course.
\nThank you,
\nYunus
Welcome! You’re on the right track. Hugging Face AutoTrain does support fine-tuning instruction-style datasets like Alpaca, but it’s a bit limited compared to manual training.
For datasets with instruction / input / output, the standard approach is to merge instruction + input into a single prompt column, and keep output as the label. AutoTrain usually expects just one “text” and one “label/output” field.
\nIf the UI only shows one mapping field, you’ll need to preprocess your dataset before uploading (e.g., combine instruction + input into a new prompt column).
For full control, many people skip AutoTrain and instead use the Hugging Face trl library (SFTTrainer) with LoRA. This gives you more flexibility for instruction-tuning LLaMA models.
Docs to check:
\nFine-tuning with TRL
\nAutoTrain docs
\nSo TL;DR: preprocess into 2 columns (prompt, output), then upload to AutoTrain, or use trl for more advanced setups.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-10-07T15:04:17.287Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168711, 'topic_slug': 'auto-train-with-alpaca-model-data-set', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/auto-train-with-alpaca-model-data-set/168711/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi there,
+I’m new both on this forum and huggingface world. Please go easy on me
+I have a question to ask. I want to use auto train for fine tune a model like meta-llama/Llama-3.1-8B-Instruct. I have a data set which is in alpaca model with instruction, input and output columns.
Questions are;
+I couldn’t find a good document or example in order to learn how to fine tune a model with using this type of model.
+None of the information buttons are working on the Auto Train screen like the one above task or parameter combo-box.
+How can I put more fields in column mapping section? There is only one right now. I think I should put instruction, input and output columns.
+If there is any good documentation, please share it with me. So, I can started to learn some stuff.
+ +Best regards,
+Yunus Emre
I don’t have any data science knowledge whatsoever, but I think we can manage if we just do some basic preprocessing in Python… Functions for data processing and shaping are usually available somewhere in the libraries.
+Use one rendered text column for SFT. Do not map instruction/input/output separately. Convert your rows to the model’s chat format, save as a single-column dataset, and map text → text in AutoTrain. (Hugging Face)
meta-llama/Llama-3.1-8B-Instruct. (Hugging Face)SFTTrainer accepts either:
+{""text"": ""...final rendered conversation...""}, or{""prompt"": ""..."", ""completion"": ""...""}.text column for chat SFT. (Hugging Face)instruction + (""\n\n"" + input if present); assistant = output.apply_chat_template(messages, tokenize=False, add_generation_prompt=False).from datasets import load_dataset
+from transformers import AutoTokenizer
+import pandas as pd
+
+tok = AutoTokenizer.from_pretrained(""meta-llama/Llama-3.1-8B-Instruct"")
+
+def render_row(r):
+ user = r[""instruction""] + ((""\n\n"" + r[""input""]) if r.get(""input"") else """")
+ messages = [{""role"":""user"",""content"":user},
+ {""role"":""assistant"",""content"":r[""output""]}]
+ return tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=False)
+
+ds = load_dataset(""tatsu-lab/alpaca"", split=""train"") # replace with your data
+df = pd.DataFrame({""text"": [render_row(x) for x in ds]})
+df.to_csv(""autotrain_llm_sft.csv"", index=False)
+apply_chat_template ensures the exact prompt tokens and headers the model expects. (Hugging Face)
UI: upload CSV/JSONL, set Column Mapping → text → text, choose LLM finetuning → SFT. (Hugging Face)
+CLI (reliable, explicit):
pip install autotrain-advanced
+
+autotrain llm \
+ --train \
+ --project-name llama31-alpaca-sft \
+ --model meta-llama/Llama-3.1-8B-Instruct \
+ --data-path ./ \
+ --train-split train \
+ --text-column text \
+ --trainer sft \
+ --use-peft \
+ --lora-r 16 --lora-alpha 32 --lora-dropout 0.05 \
+ --batch-size 4 --gradient-accumulation 8 \
+ --lr 2e-4 --epochs 3 --bf16 \
+ --max-seq-length 4096
+Flags mirror documented AutoTrain usage. Adjust batch and GA for VRAM. (Hugging Face)
+At generation, build messages and call the same tokenizer’s chat template to format the prompt before generate. Template mismatches degrade outputs. Llama 3.1 has known header nuances; verify your output. (Hugging Face)
Only if you pick a different trainer or format:
+prompt and completion. (Hugging Face)prompt, chosen, rejected. AutoTrain exposes those roles in column mapping. (Hugging Face)--text-column text. (Hugging Face)AutoTrain LLM finetuning and column mapping, TRL SFT dataset formats, and chat templating docs. (Hugging Face)
" +All my spaces are down after rebuild,https://discuss.huggingface.co/t/all-my-spaces-are-down-after-rebuild/168915,168915,24,2025-10-05 04:59:57.954000+00:00,"[{'id': 243077, 'name': 'Winston', 'username': 'winstxnhdw', 'avatar_template': '/user_avatar/discuss.huggingface.co/winstxnhdw/{size}/29933_2.png', 'created_at': '2025-10-05T04:59:58.011Z', 'cooked': 'According to my logs on Grafana, they’ve been down since 2025-10-05 02:40:46 +0000
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-10-05T04:59:58.011Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 11, 'readers_count': 10, 'score': 17.2, 'yours': False, 'topic_id': 168915, 'topic_slug': 'all-my-spaces-are-down-after-rebuild', 'display_username': 'Winston', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 29343, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/all-my-spaces-are-down-after-rebuild/168915/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 243078, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-10-05T05:39:10.176Z', 'cooked': 'Did you make any changes to the Docker image? If not, this case might be similar…
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-10-05T05:39:10.176Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 11, 'readers_count': 10, 'score': 2.2, 'yours': False, 'topic_id': 168915, 'topic_slug': 'all-my-spaces-are-down-after-rebuild', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/streamlit-docker-space-permanently-in-building-state/168910/3', 'internal': True, 'reflection': False, 'title': 'Streamlit Docker space permanently in ""Building"" state', 'clicks': 6}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/all-my-spaces-are-down-after-rebuild/168915/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 243091, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-05T17:39:29.308Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-10-05T17:39:29.308Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 168915, 'topic_slug': 'all-my-spaces-are-down-after-rebuild', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/all-my-spaces-are-down-after-rebuild/168915/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","According to my logs on Grafana, they’ve been down since 2025-10-05 02:40:46 +0000
","Did you make any changes to the Docker image? If not, this case might be similar…
" +"Qwen Image, ComfyUI and Python Script",https://discuss.huggingface.co/t/qwen-image-comfyui-and-python-script/168684,168684,5,2025-09-25 20:23:15.694000+00:00,"[{'id': 242583, 'name': 'Bo Andersen', 'username': 'boan-dk', 'avatar_template': '/user_avatar/discuss.huggingface.co/boan-dk/{size}/54270_2.png', 'created_at': '2025-09-25T20:23:15.760Z', 'cooked': 'I am wondering what ComfyUI are doing with the models (e.g. Qwen Image). They can run on consumer hardware where the official seems to use at lot more resources.
\nI have tried to use the script from Qwen/Qwen-Image · Hugging Face and changed the model to Comfy-Org/Qwen-Image_ComfyUI · Hugging Face
\nIt seems they are two different formats/packages. Can anyone suggest a refactored script that works with the ComfyUI model?
\nThanks
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-25T20:50:09.655Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 4, 'readers_count': 3, 'score': 65.8, 'yours': False, 'topic_id': 168684, 'topic_slug': 'qwen-image-comfyui-and-python-script', 'display_username': 'Bo Andersen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI', 'internal': False, 'reflection': False, 'title': 'Comfy-Org/Qwen-Image_ComfyUI · Hugging Face', 'clicks': 2}, {'url': 'https://huggingface.co/Qwen/Qwen-Image', 'internal': False, 'reflection': False, 'title': 'Qwen/Qwen-Image · Hugging Face', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104489, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/qwen-image-comfyui-and-python-script/168684/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242602, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-25T23:24:29.782Z', 'cooked': 'ComfyUI and Diffusers are entirely different software, so conversion isn’t really something you should consider. It’s not impossible, but most models have weights for both software available on Hugging Face, so use the weights provided there…
\nThere are ways to use ComfyUI via its API. Also, when using Diffusers, while the sample scripts prioritize accuracy and code simplicity, there are methods for memory optimization and speeding up the process in actual use.
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-25T23:24:29.782Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 168684, 'topic_slug': 'qwen-image-comfyui-and-python-script', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/qwen_image_comfy_diffusers_python.md', 'internal': False, 'reflection': False, 'title': 'qwen_image_comfy_diffusers_python.md · John6666/forum1 at main', 'clicks': 4}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/qwen-image-comfyui-and-python-script/168684/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242612, 'name': 'Bo Andersen', 'username': 'boan-dk', 'avatar_template': '/user_avatar/discuss.huggingface.co/boan-dk/{size}/54270_2.png', 'created_at': '2025-09-26T05:01:12.123Z', 'cooked': '\n\nmost models have weights for both software available on Hugging Face
\n
Can you provide a link for the weights to a model where I can see the differences for both software?
\nThank you
\n\nthe weights to a model where I can see the differences for both software
\n
Qwen/Qwen-Image vs Comfy-Org/Qwen-Image_ComfyUI is also an example…
stabilityai/stable-diffusion-xl-base-1.0
\n
safetensors files are not simply split and merged; the keys have changed. While conversion is possible (The actual conversion method varies depending on the model architecture.), it’s best to avoid it if you’re unsure. It’s best to use files intended for ComfyUI with ComfyUI, and files intended for Diffusers with Diffusers.', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-26T06:15:30.478Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 168684, 'topic_slug': 'qwen-image-comfyui-and-python-script', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0', 'internal': False, 'reflection': False, 'title': 'stabilityai/stable-diffusion-xl-base-1.0 · Hugging Face', 'clicks': 0}, {'url': 'https://github.com/huggingface/diffusers/blob/main/scripts/convert_diffusers_to_original_sdxl.py', 'internal': False, 'reflection': False, 'title': 'diffusers/scripts/convert_diffusers_to_original_sdxl.py at main · huggingface/diffusers · GitHub', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/qwen-image-comfyui-and-python-script/168684/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 243088, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-05T15:33:40.629Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-10-05T15:33:40.629Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168684, 'topic_slug': 'qwen-image-comfyui-and-python-script', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/qwen-image-comfyui-and-python-script/168684/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am wondering what ComfyUI are doing with the models (e.g. Qwen Image). They can run on consumer hardware where the official seems to use at lot more resources.
+I have tried to use the script from Qwen/Qwen-Image · Hugging Face and changed the model to Comfy-Org/Qwen-Image_ComfyUI · Hugging Face
+It seems they are two different formats/packages. Can anyone suggest a refactored script that works with the ComfyUI model?
+Thanks
","ComfyUI and Diffusers are entirely different software, so conversion isn’t really something you should consider. It’s not impossible, but most models have weights for both software available on Hugging Face, so use the weights provided there…
+There are ways to use ComfyUI via its API. Also, when using Diffusers, while the sample scripts prioritize accuracy and code simplicity, there are methods for memory optimization and speeding up the process in actual use.
" +Help: Can’t find Multi Image Input node in ComfyUI,https://discuss.huggingface.co/t/help-can-t-find-multi-image-input-node-in-comfyui/168826,168826,5,2025-10-01 08:10:20.352000+00:00,"[{'id': 242889, 'name': 'yaoyuan', 'username': 'graceyaoyuan', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/g/bbce88/{size}.png', 'created_at': '2025-10-01T08:10:20.415Z', 'cooked': 'Hi everyone,
\nI uploaded a workflow in ComfyUI, but it shows that a Multi Image Input node is missing.
\nI don’t know where to download this node or how to fix the issue.
\nHas anyone encountered this before, or can point me in the right direction? Thanks!
I’m not a ComfyUI user, so I can’t be certain, but it looks like you’ll need either ComfyUI_pixtral_vision or ComfyUI Pixtral Large Extension…?
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-01T08:36:09.112Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 168826, 'topic_slug': 'help-can-t-find-multi-image-input-node-in-comfyui', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/ShmuelRonen/ComfyUI_pixtral_vision', 'internal': False, 'reflection': False, 'title': 'GitHub - ShmuelRonen/ComfyUI_pixtral_vision: The `ComfyUI_pixtral_vision` node is a powerful ComfyUI node designed to integrate seamlessly with the Mistral Pixtral API. It facilitates the analysis of images through deep learning models, interpreting and d', 'clicks': 1}, {'url': 'https://github.com/ShmuelRonen/ComfyUI_pixtral_large', 'internal': False, 'reflection': False, 'title': ""GitHub - ShmuelRonen/ComfyUI_pixtral_large: A ComfyUI custom node that integrates Mistral AI's Pixtral Large vision model, enabling powerful multimodal AI capabilities within ComfyUI. Pixtral Large is a 124B parameter model (123B decoder + 1B vision encod"", 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/help-can-t-find-multi-image-input-node-in-comfyui/168826/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242959, 'name': 'yaoyuan', 'username': 'graceyaoyuan', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/g/bbce88/{size}.png', 'created_at': '2025-10-02T01:11:40.507Z', 'cooked': 'Hi John,
\nThanks so much! I downloaded the ComfyUI_pixtral_vision and it works — no more red alerts.
I can’t believe you’re not a ComfyUI user; you seem like a master!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-02T13:12:34.049Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168826, 'topic_slug': 'help-can-t-find-multi-image-input-node-in-comfyui', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/help-can-t-find-multi-image-input-node-in-comfyui/168826/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I uploaded a workflow in ComfyUI, but it shows that a Multi Image Input node is missing.
+I don’t know where to download this node or how to fix the issue.
+Has anyone encountered this before, or can point me in the right direction? Thanks!
I’m not a ComfyUI user, so I can’t be certain, but it looks like you’ll need either ComfyUI_pixtral_vision or ComfyUI Pixtral Large Extension…?
" +Request to reset paper authorship,https://discuss.huggingface.co/t/request-to-reset-paper-authorship/168822,168822,5,2025-10-01 02:01:48.922000+00:00,"[{'id': 242881, 'name': 'Zixin Zhu', 'username': 'buxiangzhiren', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/97f17d/{size}.png', 'created_at': '2025-10-01T02:01:48.980Z', 'cooked': 'Hi HF team,
\nI’m the author of the following arXiv papers (due to link limits, I’m listing only one here), but on my Hugging Face profile the authorship appears to be claimed by a different account (or my claim stays pending due to a conflict). Could you please help reset/transfer the claim to my main account?
\n\nThanks a lot!
\nBest,
\nZixin
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-01T02:01:48.980Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 168822, 'topic_slug': 'request-to-reset-paper-authorship', 'display_username': 'Zixin Zhu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/papers/2306.04632', 'internal': False, 'reflection': False, 'title': 'Paper page - Designing a Better Asymmetric VQGAN for StableDiffusion', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104804, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/request-to-reset-paper-authorship/168822/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242884, 'name': 'hysts', 'username': 'hysts', 'avatar_template': '/user_avatar/discuss.huggingface.co/hysts/{size}/32230_2.png', 'created_at': '2025-10-01T03:53:44.972Z', 'cooked': 'Hi @buxiangzhiren ,
\nThanks for reporting this, and sorry for the trouble. I’ve shared this internally, and the team will look into it.
The issue should be resolved now. Thanks again for reporting it.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-01T10:31:05.129Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168822, 'topic_slug': 'request-to-reset-paper-authorship', 'display_username': 'hysts', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 7263, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/request-to-reset-paper-authorship/168822/3', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242931, 'name': 'Zixin Zhu', 'username': 'buxiangzhiren', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/97f17d/{size}.png', 'created_at': '2025-10-01T21:36:29.249Z', 'cooked': 'Hi @hysts , thank you for your help!
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-10-01T21:36:29.249Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168822, 'topic_slug': 'request-to-reset-paper-authorship', 'display_username': 'Zixin Zhu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104804, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/request-to-reset-paper-authorship/168822/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242980, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-02T09:36:48.064Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-10-02T09:36:48.064Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168822, 'topic_slug': 'request-to-reset-paper-authorship', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/request-to-reset-paper-authorship/168822/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi HF team,
+I’m the author of the following arXiv papers (due to link limits, I’m listing only one here), but on my Hugging Face profile the authorship appears to be claimed by a different account (or my claim stays pending due to a conflict). Could you please help reset/transfer the claim to my main account?
+ +Thanks a lot!
+Best,
+Zixin
",The issue should be resolved now. Thanks again for reporting it.
+"Is it possible to remove articles (the, a, an) from a text sample without consequences?",https://discuss.huggingface.co/t/is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences/168801,168801,10,2025-09-30 09:20:23.391000+00:00,"[{'id': 242835, 'name': 'CockroachTraveler', 'username': 'CockroachTraveler', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/b5ac83/{size}.png', 'created_at': '2025-09-30T09:20:23.450Z', 'cooked': 'In my experience, these articles do not make significant sense, but they take up some amount of data.
\nActually, the crux of the question is, if they are previously removed from the text selection, will this reduce costs and will this not affect the perception of the meaning of the test by the model?
(task: text generation or text2image Lora)
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-30T09:22:48.663Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 6, 'readers_count': 5, 'score': 46.2, 'yours': False, 'topic_id': 168801, 'topic_slug': 'is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences', 'display_username': 'CockroachTraveler', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 62158, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences/168801/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242866, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-30T21:15:23.799Z', 'cooked': 'This should generally be avoided as it significantly impacts output results. However, it is possible to train models to omit articles, and while rare, I have seen examples. Naturally, this comes at a higher cost.
\nLet’s just use it as is…
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-30T21:15:23.799Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 21.2, 'yours': False, 'topic_id': 168801, 'topic_slug': 'is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/removing_articles_affect_results.md', 'internal': False, 'reflection': False, 'title': 'removing_articles_affect_results.md · John6666/forum1 at main', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences/168801/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242890, 'name': 'CockroachTraveler', 'username': 'CockroachTraveler', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/b5ac83/{size}.png', 'created_at': '2025-10-01T08:26:07.022Z', 'cooked': 'Thanks for the reply, although sad. However, I would like to clarify which tests you used to state this.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-10-01T09:18:31.408Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 6, 'readers_count': 5, 'score': 26.2, 'yours': False, 'topic_id': 168801, 'topic_slug': 'is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences', 'display_username': 'CockroachTraveler', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 62158, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences/168801/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242929, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-10-01T20:27:00.088Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-01T20:27:00.088Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 168801, 'topic_slug': 'is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/is-it-possible-to-remove-articles-the-a-an-from-a-text-sample-without-consequences/168801/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","In my experience, these articles do not make significant sense, but they take up some amount of data.
+Actually, the crux of the question is, if they are previously removed from the text selection, will this reduce costs and will this not affect the perception of the meaning of the test by the model?
(task: text generation or text2image Lora)
","This should generally be avoided as it significantly impacts output results. However, it is possible to train models to omit articles, and while rare, I have seen examples. Naturally, this comes at a higher cost.
+Let’s just use it as is…
" +KeyError: ‘classifier.dense.weight’ when loading LoRA adapter with quantized Roberta classification model,https://discuss.huggingface.co/t/keyerror-classifier-dense-weight-when-loading-lora-adapter-with-quantized-roberta-classification-model/168793,168793,5,2025-09-30 01:27:54.577000+00:00,"[{'id': 242812, 'name': 'AkiraNom', 'username': 'TetorisAce', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/t/74df32/{size}.png', 'created_at': '2025-09-30T01:27:54.639Z', 'cooked': 'Hi all,
\nI fine-tuned a quantized roberta-base classification model using PEFT + LoRA. Then, training runs fine, and I save the adapter.
from datasets import load_dataset\nimport evaluate\nfrom peft import (\n LoraConfig,\n TaskType,\n get_peft_model,\n prepare_model_for_kbit_training\n)\nimport torch\nfrom transformers import (\n AutoTokenizer,\n DataCollatorWithPadding,\n AutoModelForSequenceClassification,\n BitsAndBytesConfig,\n Trainer,\n TrainingArguments\n)\ncheckpoint = ""dstefa/roberta-base_topic_classification_nyt_news""\n\n# create quantization object\nquantization_config = BitsAndBytesConfig(\n load_in_4bit=True,\n bnb_4bit_quant_type=""nf4"",\n bnb_4bit_use_double_quant=True,\n bnb_4bit_compute_dtype=torch.bfloat16,\n llm_int8_skip_modules=[""classifier""] \n)\n\nbase_model = AutoModelForSequenceClassification.from_pretrained(\n checkpoint,\n num_labels=num_labels,\n id2label=id2label,\n label2id=label2id,\n ignore_mismatched_sizes=True,\n quantization_config=quantization_config\n )\n\n# preprocess the quantized model for training\nmodel = prepare_model_for_kbit_training(base_model)\n\n# create LoRA config object\nlora_config = LoraConfig(\n task_type=TaskType.SEQ_CLS,\n inference_mode=False, # set to Fasle for training\n r=8,\n lora_alpha=16,\n lora_dropout=0.1,\n bias=\'none\',\n modules_to_save=[""classifier.dense"", ""classifier.out_proj""],\n )\n\n# create a trainable PeftModel\nfinal_model = get_peft_model(model, lora_config)\n\nfinal_training_args = TrainingArguments(\n output_dir=""/content/drive/MyDrive/Projects/new-topic-classifier/checkpoint/"",\n num_train_epochs=2,\n # eval_strategy=""epoch"",\n # save_strategy=""epoch"",\n eval_strategy=""steps"", \n eval_steps=10000, \n save_strategy=""steps"", \n save_steps=10000, \n save_total_limit=3, \n load_best_model_at_end=False, \n logging_strategy=""steps"",\n logging_steps=50,\n logging_first_step=True,\n fp16=True,\n run_name=""final_topic_classifier_run"",\n report_to=""wandb"", # W&B is active\n push_to_hub=True,\n hub_model_id=""####/New-topic-classifier-training-model-storage"",\n hub_strategy=""checkpoint"",\n)\n\nfinal_trainer = Trainer(\n model=final_model,\n args=final_training_args,\n train_dataset=train_dataset,\n eval_dataset=val_dataset,\n processing_class=tokenizer,\n data_collator=data_collator,\n compute_metrics=compute_metrics,\n)\n\nfinal_trainer.train()\n\n# Save the adapter model after training\nadapter_output_dir = ""/content/drive/MyDrive/Projects/new-topic-classifier/final_adapter""\nfinal_trainer.model.save_pretrained(adapter_output_dir)\n\n# Push the adapter model to Hugging Face Hub\nadapter_repo_name = ""XXXX/agnews_classifier_naive_model_adapters""\nfinal_trainer.model.push_to_hub(adapter_repo_name)\nBut when I try to use if for inference like this
\n## inference\ncheckpoint = ""dstefa/roberta-base_topic_classification_nyt_news""\nadapter_repo_name = ""XXXX/agnews_classifier_naive_model_adapters""\n\n# create quantization object\nquantization_config = BitsAndBytesConfig(\n load_in_4bit=True,\n bnb_4bit_quant_type=""nf4"",\n bnb_4bit_use_double_quant=True,\n bnb_4bit_compute_dtype=torch.bfloat16,\n llm_int8_skip_modules=[""classifier""] \n)\n\nbase_model = AutoModelForSequenceClassification.from_pretrained(\n checkpoint,\n num_labels=num_labels,\n id2label=id2label,\n label2id=label2id,\n ignore_mismatched_sizes=True,\n quantization_config=quantization_config\n )\n\nbase_model.load_adapter(adapter_repo_name)\nI got an error:
\nKeyError: \'classifier.dense.weight\'\nI tried another way to load a model with the adapter, but it returned the same error.
\nPeftModel.from_pretrained(base_model, adapter_repo_name)\nHow should I properly load an adapter for inference in a quantized sequence classification model? Is the issue related to any config setting or training arguments?
\nThank you for your help in advance.
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-30T01:27:54.639Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 3, 'readers_count': 2, 'score': 50.6, 'yours': False, 'topic_id': 168793, 'topic_slug': 'keyerror-classifier-dense-weight-when-loading-lora-adapter-with-quantized-roberta-classification-model', 'display_username': 'AkiraNom', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104736, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/keyerror-classifier-dense-weight-when-loading-lora-adapter-with-quantized-roberta-classification-model/168793/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242813, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-30T01:48:53.309Z', 'cooked': 'save/load method deviating from PEFT’s design?
\nRoot cause: you saved submodules of the head. At load time PEFT expects the whole classification head to be in modules_to_save, not its internal layers. With 4-bit quantization this mismatch often surfaces as KeyError: \'classifier.dense.weight\'. Save modules_to_save=[""classifier""], then load the adapter into the quantized base via PeftModel.from_pretrained. (Hugging Face)
# Training change — save the entire head, not its sublayers\n# Docs: https://huggingface.co/docs/peft/en/developer_guides/troubleshooting\nlora_config = LoraConfig(\n task_type=TaskType.SEQ_CLS,\n r=8, lora_alpha=16, lora_dropout=0.1, bias=""none"",\n modules_to_save=[""classifier""], # <= change\n # Optionally specify target modules; RoBERTa attention/FFN names vary by model\n # target_modules=[""query"",""key"",""value"",""dense"",""intermediate.dense"",""output.dense""]\n)\nKey point repeated two ways:
\n""classifier"").""classifier.dense"" or ""classifier.out_proj"". (Hugging Face)# Inference — load quantized base, then attach adapter\n# BitsAndBytes: https://huggingface.co/docs/transformers/en/quantization/bitsandbytes\nfrom peft import PeftModel, PeftConfig\nfrom transformers import AutoModelForSequenceClassification, BitsAndBytesConfig\n\ncheckpoint = ""dstefa/roberta-base_topic_classification_nyt_news""\nadapter_repo = ""XXXX/agnews_classifier_naive_model_adapters""\n\nbnb = BitsAndBytesConfig(\n load_in_4bit=True,\n bnb_4bit_quant_type=""nf4"",\n bnb_4bit_use_double_quant=True,\n bnb_4bit_compute_dtype=torch.bfloat16,\n)\n\nbase = AutoModelForSequenceClassification.from_pretrained(\n checkpoint,\n num_labels=num_labels, id2label=id2label, label2id=label2id,\n quantization_config=bnb, device_map=""auto"",\n)\n\n# Keep the head in float to avoid 4-bit dtype conflicts\nbase.classifier.float()\n\n# Load adapter properly (do NOT call load_adapter on the raw base model)\n# Correct API: https://huggingface.co/docs/peft/en/developer_guides/troubleshooting\nmodel = PeftModel.from_pretrained(base, adapter_repo)\nmodel.eval()\nKey points repeated two ways:
\nPeftModel.from_pretrained(base, adapter_id) to attach the adapter.base_model.load_adapter(...) unless base_model is already a PeftModel. (Hugging Face)ignore_mismatched_sizes=True at inference. It can silently re-init a head with the wrong shape.pip install -U peft. (Hugging Face)prepare_model_for_kbit_training at inference. Use it only during training.modules_to_save=[""classifier"",""pooler""]. (Hugging Face)modules_to_save. If you pass leaf names, the wrapper mapping won’t match after quantization replaces nn.Linear with bnb.nn.Linear4bit, so PEFT can’t find classifier.dense.weight on load. Saving the whole classifier avoids that mismatch. (GitHub)modules_to_save=[""classifier""].base.classifier.float().PeftModel.from_pretrained(base, adapter_repo).model.eval() and run inference.modules_to_save guidance. (Hugging Face)Thanks for the detailed explanation—it helped a lot!
\nJust a small clarification from my side: I had to keep ignore_mismatched_sizes=True, otherwise I encountered the following error during model loading:
RuntimeError: Error(s) in loading state_dict for Linear:\n\tsize mismatch for weight: copying a param with shape torch.Size([8, 768]) from checkpoint, the shape in current model is torch.Size([14, 768]).\nSo in my case, setting ignore_mismatched_sizes=True was necessary to avoid shape mismatch issues when loading the state dict.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-10-01T12:45:26.414Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168793, 'topic_slug': 'keyerror-classifier-dense-weight-when-loading-lora-adapter-with-quantized-roberta-classification-model', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/keyerror-classifier-dense-weight-when-loading-lora-adapter-with-quantized-roberta-classification-model/168793/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi all,
+I fine-tuned a quantized roberta-base classification model using PEFT + LoRA. Then, training runs fine, and I save the adapter.
from datasets import load_dataset
+import evaluate
+from peft import (
+ LoraConfig,
+ TaskType,
+ get_peft_model,
+ prepare_model_for_kbit_training
+)
+import torch
+from transformers import (
+ AutoTokenizer,
+ DataCollatorWithPadding,
+ AutoModelForSequenceClassification,
+ BitsAndBytesConfig,
+ Trainer,
+ TrainingArguments
+)
+checkpoint = ""dstefa/roberta-base_topic_classification_nyt_news""
+
+# create quantization object
+quantization_config = BitsAndBytesConfig(
+ load_in_4bit=True,
+ bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_use_double_quant=True,
+ bnb_4bit_compute_dtype=torch.bfloat16,
+ llm_int8_skip_modules=[""classifier""]
+)
+
+base_model = AutoModelForSequenceClassification.from_pretrained(
+ checkpoint,
+ num_labels=num_labels,
+ id2label=id2label,
+ label2id=label2id,
+ ignore_mismatched_sizes=True,
+ quantization_config=quantization_config
+ )
+
+# preprocess the quantized model for training
+model = prepare_model_for_kbit_training(base_model)
+
+# create LoRA config object
+lora_config = LoraConfig(
+ task_type=TaskType.SEQ_CLS,
+ inference_mode=False, # set to Fasle for training
+ r=8,
+ lora_alpha=16,
+ lora_dropout=0.1,
+ bias='none',
+ modules_to_save=[""classifier.dense"", ""classifier.out_proj""],
+ )
+
+# create a trainable PeftModel
+final_model = get_peft_model(model, lora_config)
+
+final_training_args = TrainingArguments(
+ output_dir=""/content/drive/MyDrive/Projects/new-topic-classifier/checkpoint/"",
+ num_train_epochs=2,
+ # eval_strategy=""epoch"",
+ # save_strategy=""epoch"",
+ eval_strategy=""steps"",
+ eval_steps=10000,
+ save_strategy=""steps"",
+ save_steps=10000,
+ save_total_limit=3,
+ load_best_model_at_end=False,
+ logging_strategy=""steps"",
+ logging_steps=50,
+ logging_first_step=True,
+ fp16=True,
+ run_name=""final_topic_classifier_run"",
+ report_to=""wandb"", # W&B is active
+ push_to_hub=True,
+ hub_model_id=""####/New-topic-classifier-training-model-storage"",
+ hub_strategy=""checkpoint"",
+)
+
+final_trainer = Trainer(
+ model=final_model,
+ args=final_training_args,
+ train_dataset=train_dataset,
+ eval_dataset=val_dataset,
+ processing_class=tokenizer,
+ data_collator=data_collator,
+ compute_metrics=compute_metrics,
+)
+
+final_trainer.train()
+
+# Save the adapter model after training
+adapter_output_dir = ""/content/drive/MyDrive/Projects/new-topic-classifier/final_adapter""
+final_trainer.model.save_pretrained(adapter_output_dir)
+
+# Push the adapter model to Hugging Face Hub
+adapter_repo_name = ""XXXX/agnews_classifier_naive_model_adapters""
+final_trainer.model.push_to_hub(adapter_repo_name)
+But when I try to use if for inference like this
+## inference
+checkpoint = ""dstefa/roberta-base_topic_classification_nyt_news""
+adapter_repo_name = ""XXXX/agnews_classifier_naive_model_adapters""
+
+# create quantization object
+quantization_config = BitsAndBytesConfig(
+ load_in_4bit=True,
+ bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_use_double_quant=True,
+ bnb_4bit_compute_dtype=torch.bfloat16,
+ llm_int8_skip_modules=[""classifier""]
+)
+
+base_model = AutoModelForSequenceClassification.from_pretrained(
+ checkpoint,
+ num_labels=num_labels,
+ id2label=id2label,
+ label2id=label2id,
+ ignore_mismatched_sizes=True,
+ quantization_config=quantization_config
+ )
+
+base_model.load_adapter(adapter_repo_name)
+I got an error:
+KeyError: 'classifier.dense.weight'
+I tried another way to load a model with the adapter, but it returned the same error.
+PeftModel.from_pretrained(base_model, adapter_repo_name)
+How should I properly load an adapter for inference in a quantized sequence classification model? Is the issue related to any config setting or training arguments?
+Thank you for your help in advance.
","save/load method deviating from PEFT’s design?
+Root cause: you saved submodules of the head. At load time PEFT expects the whole classification head to be in modules_to_save, not its internal layers. With 4-bit quantization this mismatch often surfaces as KeyError: 'classifier.dense.weight'. Save modules_to_save=[""classifier""], then load the adapter into the quantized base via PeftModel.from_pretrained. (Hugging Face)
# Training change — save the entire head, not its sublayers
+# Docs: https://huggingface.co/docs/peft/en/developer_guides/troubleshooting
+lora_config = LoraConfig(
+ task_type=TaskType.SEQ_CLS,
+ r=8, lora_alpha=16, lora_dropout=0.1, bias=""none"",
+ modules_to_save=[""classifier""], # <= change
+ # Optionally specify target modules; RoBERTa attention/FFN names vary by model
+ # target_modules=[""query"",""key"",""value"",""dense"",""intermediate.dense"",""output.dense""]
+)
+Key point repeated two ways:
+""classifier"").""classifier.dense"" or ""classifier.out_proj"". (Hugging Face)# Inference — load quantized base, then attach adapter
+# BitsAndBytes: https://huggingface.co/docs/transformers/en/quantization/bitsandbytes
+from peft import PeftModel, PeftConfig
+from transformers import AutoModelForSequenceClassification, BitsAndBytesConfig
+
+checkpoint = ""dstefa/roberta-base_topic_classification_nyt_news""
+adapter_repo = ""XXXX/agnews_classifier_naive_model_adapters""
+
+bnb = BitsAndBytesConfig(
+ load_in_4bit=True,
+ bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_use_double_quant=True,
+ bnb_4bit_compute_dtype=torch.bfloat16,
+)
+
+base = AutoModelForSequenceClassification.from_pretrained(
+ checkpoint,
+ num_labels=num_labels, id2label=id2label, label2id=label2id,
+ quantization_config=bnb, device_map=""auto"",
+)
+
+# Keep the head in float to avoid 4-bit dtype conflicts
+base.classifier.float()
+
+# Load adapter properly (do NOT call load_adapter on the raw base model)
+# Correct API: https://huggingface.co/docs/peft/en/developer_guides/troubleshooting
+model = PeftModel.from_pretrained(base, adapter_repo)
+model.eval()
+Key points repeated two ways:
+PeftModel.from_pretrained(base, adapter_id) to attach the adapter.base_model.load_adapter(...) unless base_model is already a PeftModel. (Hugging Face)ignore_mismatched_sizes=True at inference. It can silently re-init a head with the wrong shape.pip install -U peft. (Hugging Face)prepare_model_for_kbit_training at inference. Use it only during training.modules_to_save=[""classifier"",""pooler""]. (Hugging Face)modules_to_save. If you pass leaf names, the wrapper mapping won’t match after quantization replaces nn.Linear with bnb.nn.Linear4bit, so PEFT can’t find classifier.dense.weight on load. Saving the whole classifier avoids that mismatch. (GitHub)modules_to_save=[""classifier""].base.classifier.float().PeftModel.from_pretrained(base, adapter_repo).model.eval() and run inference.modules_to_save guidance. (Hugging Face)I am using ImageToImageTargetSize paramenter with InferenceClient
\nfrom huggingface_hub.inference._generated.types.image_to_image import ImageToImageTargetSize
\ntarget_size=ImageToImageTargetSize(256, 256)
\nBut the output is still same as input image size. Can anyone help me to figure out what thing I am doing wrong?
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-28T07:02:20.716Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 25.6, 'yours': False, 'topic_id': 168739, 'topic_slug': 'target-size-issue', 'display_username': 'TSR', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104625, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/target-size-issue/168739/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242712, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-28T08:53:37.339Z', 'cooked': 'The parameter seems to be ignored…
\nDepending on the model, resolution constraints or the input image resolution may take precedence, causing the output resolution parameter to be ignored. Or is it a bug?
\nfrom huggingface_hub import InferenceClient, ImageToImageTargetSize\n\nclient = InferenceClient(model=""Qwen/Qwen-Image-Edit"")\nurl = ""https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/edit_homepage.jpg"" # (1312, 800)\n\nimg = client.image_to_image(\n url,\n prompt=""cinematic lighting"",\n target_size=ImageToImageTargetSize(height=256, width=256),\n provider=""fal""\n)\nprint(img.size) # (1312, 800)\nimg.save(""out.jpg"")\nI have read the full image to image inference repo files, there i find two output classes out of which ImageToImageTargetSize is defined in the main parameter class.
ImageToImageOutput is the other one which do the same functioning ig.
Here you can find it - https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/inference/\\_generated/types/image_to_image.py
\nI think it is a bug and I have reported it.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-28T09:28:46.763Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 168739, 'topic_slug': 'target-size-issue', 'display_username': 'TSR', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/inference/%5C_generated/types/image_to_image.py', 'internal': False, 'reflection': False, 'clicks': 1}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104625, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/target-size-issue/168739/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242714, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-28T10:03:37.016Z', 'cooked': 'Similar behavior was observed with prithivMLmods/Monochrome-Pencil. If the size specification parameter doesn’t work in Flux Kontext’s LoRA, then there are probably very few Endpoints that support size specification…
Could it be that parameters aren’t being passed correctly when TGI uses Diffusers as the backend…? @michellehbn
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-28T10:03:37.016Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 168739, 'topic_slug': 'target-size-issue', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/target-size-issue/168739/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242815, 'name': 'TSR', 'username': 'iam-tsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/iam-tsr/{size}/54346_2.png', 'created_at': '2025-09-30T03:55:46.433Z', 'cooked': 'The bug has been fixed and released in huggingface_hub==0.35.3
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-09-30T15:56:15.491Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168739, 'topic_slug': 'target-size-issue', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/target-size-issue/168739/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am using ImageToImageTargetSize paramenter with InferenceClient
+from huggingface_hub.inference._generated.types.image_to_image import ImageToImageTargetSize
+target_size=ImageToImageTargetSize(256, 256)
+But the output is still same as input image size. Can anyone help me to figure out what thing I am doing wrong?
","The bug has been fixed and released in huggingface_hub==0.35.3
It says
\nExit code: 1. Reason: => Database and media directory: /label-studio/data\n=> Static URL is set to: /static/\nTraceback (most recent call last):\n File ""/label-studio/.venv/bin/label-studio"", line 3, in <module>\n from label_studio.server import main\n File ""/label-studio/label_studio/server.py"", line 23, in <module>\n from label_studio.core.argparser import parse_input_args\n File ""/label-studio/label_studio/core/argparser.py"", line 5, in <module>\n from .settings.base import EXPORT_DIR\n File ""/label-studio/label_studio/core/settings/base.py"", line 470, in <module>\n os.makedirs(MEDIA_ROOT, exist_ok=True)\n File ""<frozen os>"", line 225, in makedirs\nPermissionError: [Errno 13] Permission denied: \'/label-studio/data/media\'\nWhen starting up
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-28T01:05:44.089Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 16, 'reads': 5, 'readers_count': 4, 'score': 76.0, 'yours': False, 'topic_id': 168735, 'topic_slug': 'permission-error-when-starting-a-lablestudio-space', 'display_username': 'Lin Chen you', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104613, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/permission-error-when-starting-a-lablestudio-space/168735/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242703, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-28T03:39:16.858Z', 'cooked': 'The cause is attempting to write to a directory that is not writable due to permissions. Setting the following environment variable would resolve this.
\nLABEL_STUDIO_BASE_DATA_DIR=/tmp/label-studio
\nAny directory with write permissions will work.
That error is pretty straightforward — Label Studio is trying to create its media folder but doesn’t have permission.
Here’s how to fix it:
\nCheck who owns the folder
\nls -ld /label-studio/data\n\nIf it’s owned by root, Label Studio (running as a different user) can’t write there.
Give yourself permission
\nsudo chown -R $USER:$USER /label-studio/data\n\nor if you’re running inside Docker, adjust ownership to the container user (often 1001 or label-studio).
Set writable permissions (if quick and dirty):
\nsudo chmod -R 777 /label-studio/data\n\nThis is less safe, but fine for local experiments.
\nIf Dockerized:
\nMount a local volume that’s writable:
\ndocker run -it -p 8080:8080 \\\n -v $(pwd)/mydata:/label-studio/data \\\n heartexlabs/label-studio:latest\n\nReplace $(pwd)/mydata with a folder on your machine you own.
Thanks! It worked!
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-28T10:36:56.104Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 168735, 'topic_slug': 'permission-error-when-starting-a-lablestudio-space', 'display_username': 'Lin Chen you', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104613, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/permission-error-when-starting-a-lablestudio-space/168735/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242730, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-28T22:37:38.529Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-09-28T22:37:38.529Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 1, 'readers_count': 0, 'score': 45.2, 'yours': False, 'topic_id': 168735, 'topic_slug': 'permission-error-when-starting-a-lablestudio-space', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/permission-error-when-starting-a-lablestudio-space/168735/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","It says
+Exit code: 1. Reason: => Database and media directory: /label-studio/data
+=> Static URL is set to: /static/
+Traceback (most recent call last):
+ File ""/label-studio/.venv/bin/label-studio"", line 3, in <module>
+ from label_studio.server import main
+ File ""/label-studio/label_studio/server.py"", line 23, in <module>
+ from label_studio.core.argparser import parse_input_args
+ File ""/label-studio/label_studio/core/argparser.py"", line 5, in <module>
+ from .settings.base import EXPORT_DIR
+ File ""/label-studio/label_studio/core/settings/base.py"", line 470, in <module>
+ os.makedirs(MEDIA_ROOT, exist_ok=True)
+ File ""<frozen os>"", line 225, in makedirs
+PermissionError: [Errno 13] Permission denied: '/label-studio/data/media'
+When starting up
","The cause is attempting to write to a directory that is not writable due to permissions. Setting the following environment variable would resolve this.
+LABEL_STUDIO_BASE_DATA_DIR=/tmp/label-studio
+Any directory with write permissions will work.
Hi,
\nI was training a small model just for fun when the error occured (after more 100k steps) :
\nrequests.exceptions.HTTPError: 403 Client Error: Forbidden for url: https://huggingface.co/datasets/HuggingFaceFW/fineweb-2/resolve/a8a99b128121a41b17d95901715603386f6b1daf/data/fra_Latn/train/000_00000.parquet
\nI’m wondering if I have reach some rate limits or else ? I guess it shoul failed way earlier if I was doing it wrong ?
\nI’m using it with streaming on:
\n ds_fr = load_dataset(\n ""HuggingFaceFW/fineweb-2"",\n name=""fra_Latn"",\n split=""train"",\n streaming=True\n )\nAny idea what the problem can be ?
\nThanks,
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-09-23T21:45:26.982Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 16, 'reads': 4, 'readers_count': 3, 'score': 80.8, 'yours': False, 'topic_id': 168620, 'topic_slug': '403-error-on-dataset-fineweb-2', 'display_username': 'Vincent Blazutti', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/HuggingFaceFW/fineweb-2/resolve/a8a99b128121a41b17d95901715603386f6b1daf/data/fra_Latn/train/000_00000.parquet', 'internal': False, 'reflection': False, 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104363, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/403-error-on-dataset-fineweb-2/168620/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242455, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-24T00:35:14.602Z', 'cooked': '\n\nHTTPError: 403 Client Error: Forbidden for url
\n
When streaming=True, shards are fetched on-demand, so it’s not unusual for errors to occur midway through fetching. Judging from the error message, it appears to be a CDN or network error, so I don’t think it’s a code issue.
Since the retry limit is likely less restrictive during login, how about doing huggingface_hub.login() beforehand during training and configuring datasets settings like increasing the retry count to enhance error tolerance?
Although I don’t think it’s the case this time, it’s not unheard of for the dataset repository to be updated while streaming the dataset—a rare scenario. To avoid this, explicitly specifying the revision would be the surest way.
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-09-24T00:37:14.134Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 168620, 'topic_slug': '403-error-on-dataset-fineweb-2', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/datasets/issues/6172', 'internal': False, 'reflection': False, 'title': 'Make Dataset streaming queries retryable · Issue #6172 · huggingface/datasets · GitHub', 'clicks': 0}, {'url': 'https://huggingface.co/docs/huggingface_hub/main/quick-start#authentication', 'internal': False, 'reflection': False, 'title': 'Quickstart', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/403-error-on-dataset-fineweb-2/168620/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242687, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-27T14:06:23.770Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-09-27T14:06:23.770Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168620, 'topic_slug': '403-error-on-dataset-fineweb-2', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/403-error-on-dataset-fineweb-2/168620/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I was training a small model just for fun when the error occured (after more 100k steps) :
+requests.exceptions.HTTPError: 403 Client Error: Forbidden for url: https://huggingface.co/datasets/HuggingFaceFW/fineweb-2/resolve/a8a99b128121a41b17d95901715603386f6b1daf/data/fra_Latn/train/000_00000.parquet
+I’m wondering if I have reach some rate limits or else ? I guess it shoul failed way earlier if I was doing it wrong ?
+I’m using it with streaming on:
+ ds_fr = load_dataset(
+ ""HuggingFaceFW/fineweb-2"",
+ name=""fra_Latn"",
+ split=""train"",
+ streaming=True
+ )
+Any idea what the problem can be ?
+Thanks,
","++HTTPError: 403 Client Error: Forbidden for url
+
When streaming=True, shards are fetched on-demand, so it’s not unusual for errors to occur midway through fetching. Judging from the error message, it appears to be a CDN or network error, so I don’t think it’s a code issue.
Since the retry limit is likely less restrictive during login, how about doing huggingface_hub.login() beforehand during training and configuring datasets settings like increasing the retry count to enhance error tolerance?
Although I don’t think it’s the case this time, it’s not unheard of for the dataset repository to be updated while streaming the dataset—a rare scenario. To avoid this, explicitly specifying the revision would be the surest way.
" +How to build a tokenizer from a vocab subset of a BPE tokenizer,https://discuss.huggingface.co/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698,168698,5,2025-09-26 08:13:16.730000+00:00,"[{'id': 242619, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-09-26T08:13:16.792Z', 'cooked': 'Hi community,
\nI want to distill a pretrained BPE tokenizer for my domain-specific corpus, is there anything to pay attention to?
\nWhat I will do in my mind is use the pretrained one to first tokenize all sentences of the corpus(I already did), find out the used token and get rid of the unused ones from the vocabulary. Should I also take care of the merges and make the new tokenizer again a BPE tokenizer or should I just use the subset of vocabulary to make a WordLevel tokenizer? Does anyone have already done the same thing?
Thanks!
\nalephpi
', 'post_number': 1, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T08:16:39.102Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 38, 'reads': 8, 'readers_count': 7, 'score': 66.6, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242625, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-26T09:09:50.549Z', 'cooked': 'It seems more stable to avoid modifying the existing BPE tokenizer as much as possible. Well, maybe because the core part of the Tokenizer library is written in Rust…
', 'post_number': 2, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T09:09:50.549Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 6, 'readers_count': 5, 'score': 31.2, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/bpe_vocab_subset.md', 'internal': False, 'reflection': False, 'title': 'bpe_vocab_subset.md · John6666/forum1 at main', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242626, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-09-26T09:36:55.003Z', 'cooked': 'I see, let me check your solution, since I really need to distill the vocabulary as it will enormously save my model size(from 50000 to <1000)
', 'post_number': 3, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T09:42:13.205Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 6, 'readers_count': 5, 'score': 26.2, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242627, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-26T09:55:08.816Z', 'cooked': 'Unless we change it to the WordLevel tokenizer, the distillation itself seems possible without affecting the Rust-written parts.
', 'post_number': 4, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T09:55:08.816Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 11.2, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/bpe_distill.md', 'internal': False, 'reflection': False, 'title': 'bpe_distill.md · John6666/forum1 at main', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242639, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-09-26T17:09:02.796Z', 'cooked': 'Hi John, I’m following your pruning script. It can be constructed and loaded, but the new tokenizer doesn’t have the same behavior as the original one, especially for merged tokens(original one merged but the new one doesn’t)
\nIs there a debug mode that we can find out how the token is merged during the tokenizer process?
', 'post_number': 5, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T17:14:57.044Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/bpe_distill.md#1-prune--rebuild-a-bpe-tokenizer-from-a-kept-token-list', 'internal': False, 'reflection': False, 'title': 'bpe_distill.md · John6666/forum1 at main', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/5', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242641, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-09-26T17:23:42.275Z', 'cooked': 'I see, there are some nuances about the merging procedure. In my case I have f,r,a,c,frac as tokens. But I don’t have any merge paths from f,r,a,c to frac since none of the intermediate combinations exists in my keep vocab file
', 'post_number': 6, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T17:23:42.275Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242643, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-09-26T21:24:34.330Z', 'cooked': 'Ah ha, I find out a way to include the minimal merge closure for all my keep vocab can be merged to, just slightly modify the function below, and I’ve validated such closure would provide exactly same behavior as the original one(at least on my corpus)
\ndef filter_merges_to_subset(merges: list[tuple[str,str]], keep: set[str]):\n # Keep merge (a,b) when (a+b) belongs to keep and join the a,b to keep to provide an accessible merge path to (a+b)\n # update the keep until no more merge paths can be found\n # BPE merges are greedy and ordered; preserve order.\n filtered_raw = []\n new_keep: Set[str] = set()\n while True:\n keep |= new_keep\n for a, b in merges:\n merged = a + b\n if merged in keep:\n if (a,b) not in filtered_raw:\n filtered_raw.append((a,b))\n new_keep.update((a,b))\n if new_keep - keep == set():\n break\n\n # reorder the filtered merges to preserve order as the raw will break the order as we add merges in multiple loops\n filtered = []\n for merge in merges:\n if merge in filtered_raw:\n filtered.append(merge)\n return filtered\nTo give some impression:
\nBefore debugging: ~950 tokens + 741 merges
\nAfter debugging: 1264 tokens + 1004 merges (some intermediate tokens for merge paths are added, though no occurrence at the end of tokenization)
\nOriginal: 50000 tokens + 49721 merges
\nBut after all, it worths distilling.
\n(Refined a little bit, the previous version worked but contains repetitive merges)
', 'post_number': 7, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T22:03:34.200Z', 'reply_count': 1, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 5, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 104516, 'username': 'alephpi', 'name': 'Sicheng Mao', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242644, 'name': 'Sicheng Mao', 'username': 'alephpi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png', 'created_at': '2025-09-26T21:33:13.215Z', 'cooked': 'BTW, thank you so much for your very detailed answer. I’m so grateful that you add so much references, would you give me a reading list that I can learn Transformers or Tokenizers? I saw you refer to a Transformers notebook blog, but perhaps you know helpful materials more than that? Sometimes I just find the chat-AIs are not so intelligent when I ask about the Transformers/Tokenizers APIs.
', 'post_number': 8, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T21:33:13.215Z', 'reply_count': 0, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'Sicheng Mao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 104516, 'username': 'alephpi', 'name': 'Sicheng Mao', 'avatar_template': '/user_avatar/discuss.huggingface.co/alephpi/{size}/54288_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104516, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/8', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242645, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-26T22:09:34.295Z', 'cooked': '\n\nI saw you refer to a Transformers notebook blog, but perhaps you know helpful materials more than that?
\n
About Transformers…
\nby Me.
by GPT.
\nTokenizers quicktour. Build and train BPE end-to-end; inspect tokenizer.json. (Hugging Face)
Transformers tokenizer API. Fast vs. slow, specials, saving, resizing. (Hugging Face)
\nLLM Course: train a new tokenizer from an old one (train_new_from_iterator). (Hugging Face)
Transformers quicktour for full workflow context. (Hugging Face)
\nYour earlier outline, consolidated.
\n“Tokenizer shrinking recipes.” Multiple working scripts and caveats. (Hugging Face Forums)
\nRemoving tokens from GPT/BPE tokenizers: why simple deletion fails; recreate backend. (Hugging Face Forums)
\nTokenizers issue on vocab reduction pitfalls and current guidance. (GitHub)
\nTrim down SentencePiece vocabulary by editing ModelProto.pieces (step-by-step). (Hugging Face)
SentencePiece training options, including hard_vocab_limit.
Summary of tokenizers: BPE vs WordPiece vs Unigram, pros and trade-offs. (Hugging Face)
\nFast tokenizers docs: offsets, alignment, performance notes. (Hugging Face)
\nBuilding a tokenizer from scratch (mix and match normalizers, pre-tokenizers, models). (Hugging Face)
\nCleaning or changing ByteLevel BPE alphabets alters coverage; know consequences. (Hugging Face Forums)
\nKeep config.vocab_size synced when resizing embeddings; common failure mode. (Hugging Face)
Space handling in BPE tokenizers (add_prefix_space) affects segmentation. (Hugging Face Forums)
Use fast tokenizers; confirm is_fast; batch properly; multiprocessing guidance. (Hugging Face Forums)
Tokenizers Python docs for API surface and saving formats. (Hugging Face)
\nUse order: quicktour → tokenizer API → LLM course train-new → shrinking threads/issues → SP trimming if Unigram → pitfalls/perf → BPE-Knockout.
', 'post_number': 9, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-09-26T23:11:23.390Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 60.6, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/ArturoNereu/AI-Study-Group', 'internal': False, 'reflection': False, 'title': 'GitHub - ArturoNereu/AI-Study-Group: Resources to learn AI', 'clicks': 1}, {'url': 'https://github.com/NielsRogge/Transformers-Tutorials', 'internal': False, 'reflection': False, 'title': 'GitHub - NielsRogge/Transformers-Tutorials: This repository contains demos I made with the Transformers library by HuggingFace.', 'clicks': 1}, {'url': 'https://discuss.huggingface.co/t/tokenizer-shrinking-recipes/8564', 'internal': True, 'reflection': False, 'title': 'Tokenizer shrinking recipes', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/en/fast_tokenizers', 'internal': False, 'reflection': False, 'title': 'Tokenizers', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/removing-tokens-from-the-gpt-tokenizer/30753', 'internal': True, 'reflection': False, 'title': 'Removing tokens from the GPT tokenizer', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/how-to-properly-clean-vocabulary-from-bbpe-tokenizer/22827', 'internal': True, 'reflection': False, 'title': 'How to properly clean vocabulary from BBPE tokenizer', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/v4.25.1/quicktour', 'internal': False, 'reflection': False, 'title': 'Quick tour', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/bpe-tokenizers-and-spaces-before-words/475', 'internal': True, 'reflection': False, 'title': 'BPE tokenizers and spaces before words', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/tokenizer-dataset-is-very-slow/19722', 'internal': True, 'reflection': False, 'title': 'Tokenizer dataset is very slow', 'clicks': 0}, {'url': 'https://huggingface.co/docs/tokenizers/python/latest/index.html', 'internal': False, 'reflection': False, 'title': 'Tokenizers — tokenizers documentation', 'clicks': 0}, {'url': 'https://huggingface.co/posts/burtenshaw/724732252831042', 'internal': False, 'reflection': False, 'title': '@burtenshaw on Hugging Face: ""new smol course If you’re building with or learning about post training AI…""', 'clicks': 0}, {'url': 'https://huggingface.co/blog/mlabonne/llm-course', 'internal': False, 'reflection': False, 'title': 'The Large Language Model Course', 'clicks': 0}, {'url': 'https://huggingface.co/learn/llm-course/en/chapter6/2', 'internal': False, 'reflection': False, 'title': 'Training a new tokenizer from an old one - Hugging Face LLM Course', 'clicks': 0}, {'url': 'https://huggingface.co/docs/tokenizers/en/quicktour', 'internal': False, 'reflection': False, 'title': 'Quicktour', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/en/tokenizer_summary', 'internal': False, 'reflection': False, 'title': 'Summary of the tokenizers', 'clicks': 0}, {'url': 'https://huggingface.co/learn/llm-course/en/chapter6/8', 'internal': False, 'reflection': False, 'title': 'Building a tokenizer, block by block - Hugging Face LLM Course', 'clicks': 0}, {'url': 'https://triton-lang.org/main/getting-started/tutorials/index.html', 'internal': False, 'reflection': False, 'title': 'Tutorials — Triton documentation', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/en/main_classes/tokenizer', 'internal': False, 'reflection': False, 'title': 'Tokenizer', 'clicks': 0}, {'url': 'https://ahmadosman.com/blog/learn-llms-roadmap/', 'internal': False, 'reflection': False, 'title': ""So You Want to Learn LLMs? Here's the Roadmap : A Real-World, No-Bloat Guide to Building, Training, and Shipping LLMs · Osman's Odyssey: Byte & Build"", 'clicks': 0}, {'url': 'https://github.com/huggingface/tokenizers/issues/1686', 'internal': False, 'reflection': False, 'title': 'Question: Shrinking Tokenizer Vocabulary for Reduced Memory Consumption with Pre-Trained Model (LLaMA) Fine-Tuning · Issue #1686 · huggingface/tokenizers · GitHub', 'clicks': 0}, {'url': 'https://huggingface.co/docs/transformers/en/quicktour', 'internal': False, 'reflection': False, 'title': 'Quickstart', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/9', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242677, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-27T10:10:11.632Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 10, 'post_type': 3, 'posts_count': 10, 'updated_at': '2025-09-27T10:10:11.632Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 2, 'readers_count': 1, 'score': 10.4, 'yours': False, 'topic_id': 168698, 'topic_slug': 'how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-build-a-tokenizer-from-a-vocab-subset-of-a-bpe-tokenizer/168698/10', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi community,
+I want to distill a pretrained BPE tokenizer for my domain-specific corpus, is there anything to pay attention to?
+What I will do in my mind is use the pretrained one to first tokenize all sentences of the corpus(I already did), find out the used token and get rid of the unused ones from the vocabulary. Should I also take care of the merges and make the new tokenizer again a BPE tokenizer or should I just use the subset of vocabulary to make a WordLevel tokenizer? Does anyone have already done the same thing?
Thanks!
+alephpi
","Unless we change it to the WordLevel tokenizer, the distillation itself seems possible without affecting the Rust-written parts.
" +Dataset Page is Crashing,https://discuss.huggingface.co/t/dataset-page-is-crashing/168659,168659,10,2025-09-25 00:35:34.612000+00:00,"[{'id': 242531, 'name': 'Andrew Drozdov', 'username': 'mrdrozdov', 'avatar_template': '/user_avatar/discuss.huggingface.co/mrdrozdov/{size}/2692_2.png', 'created_at': '2025-09-25T00:35:34.674Z', 'cooked': 'Not sure why this page is crashing. Maybe disable viewer for now? jfkback/crumb · Datasets at Hugging Face
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-25T00:35:34.674Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 8, 'readers_count': 7, 'score': 31.4, 'yours': False, 'topic_id': 168659, 'topic_slug': 'dataset-page-is-crashing', 'display_username': 'Andrew Drozdov', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/jfkback/crumb', 'internal': False, 'reflection': False, 'title': 'jfkback/crumb · Datasets at Hugging Face', 'clicks': 1}], 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4300, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-page-is-crashing/168659/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242533, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-25T00:38:37.759Z', 'cooked': 'Hmm…? Seems working for me.
\n
This is the default split. Are you able to open any of the others?
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-25T01:38:59.860Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.2, 'yours': False, 'topic_id': 168659, 'topic_slug': 'dataset-page-is-crashing', 'display_username': 'Andrew Drozdov', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4300, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-page-is-crashing/168659/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242543, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-25T05:27:15.867Z', 'cooked': 'Seems I can open them?
\n
Wow. Magically seems to work when I open incognito. No idea why. Tried disabling a bunch of extensions, but still only works in incognito. Thank you for the follow up!
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-25T13:26:10.606Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 1, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 15.8, 'yours': False, 'topic_id': 168659, 'topic_slug': 'dataset-page-is-crashing', 'display_username': 'Andrew Drozdov', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': '', 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4300, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-page-is-crashing/168659/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242609, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-26T01:27:03.999Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-09-26T01:27:03.999Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.4, 'yours': False, 'topic_id': 168659, 'topic_slug': 'dataset-page-is-crashing', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/dataset-page-is-crashing/168659/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Not sure why this page is crashing. Maybe disable viewer for now? jfkback/crumb · Datasets at Hugging Face
"," +Wow. Magically seems to work when I open incognito. No idea why. Tried disabling a bunch of extensions, but still only works in incognito. Thank you for the follow up!
" +RuntimeError: Backward through graph with Whisper-medium and gradient_checkpointing=True,https://discuss.huggingface.co/t/runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true/168571,168571,9,2025-09-21 22:04:06.519000+00:00,"[{'id': 242354, 'name': 'Brian', 'username': 'brianko', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/3da27b/{size}.png', 'created_at': '2025-09-21T22:04:06.595Z', 'cooked': 'I am trying to fine-tune Whisper-medium and am getting this specific error during trainer.train():
tmp/ipython-input-774985985.py:8: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `Seq2SeqTrainer.__init__`. Use `processing_class` instead.\n trainer = Seq2SeqTrainer(\n---------------------------------------------------------------------------\nRuntimeError Traceback (most recent call last)\n/tmp/ipython-input-774985985.py in <cell line: 0>()\n 16 tokenizer=processor,\n 17 )\n---> 18 trainer.train()\n 19 #trainer.push_to_hub()\n\n10 frames\n/usr/local/lib/python3.12/dist-packages/torch/autograd/graph.py in _engine_run_backward(t_outputs, *args, **kwargs)\n 827 unregister_hooks = _register_logging_hooks_on_whole_graph(t_outputs)\n 828 try:\n--> 829 return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass\n 830 t_outputs, *args, **kwargs\n 831 ) # Calls into the C++ engine to run the backward pass\n\nRuntimeError: Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). Saved intermediate values of the graph are freed when you call .backward() or autograd.grad(). Specify retain_graph=True if you need to backward through the graph a second time or if you need to access saved tensors after calling backward.\nThese are the steps I’ve tried:
\nGradient checkpointing enabled (gradient_checkpointing=True).
FP16 disabled (fp16=False).
use_cache=False (which is the default for training with checkpointing, but you can mention you checked).
predict_with_generate=True.
Running on a minimal dataset subset.
\nUsing the original openai/whisper-medium model.
Restarting the runtime.
\nEnv:
\nPyTorch version: 2.8.0+cu126\nTransformers version: 4.56.2\nAccelerate version: 1.10.1\nDatasets version: 4.1.1\nModified code (per Gemini):
\nfrom transformers import WhisperForConditionalGeneration\n# Diag\nfrom accelerate import Accelerator\naccelerator = Accelerator()\ndevice = accelerator.device\n\nmodel = WhisperForConditionalGeneration.from_pretrained(""openai/whisper-medium"")\n\n#Diag\nmodel.to(device)\n\nfrom functools import partial\n\n# disable cache during training since it\'s incompatible with gradient checkpointing\nmodel.config.use_cache = False\n\n# set language and task for generation and re-enable cache\nmodel.generate = partial(\n model.generate, language=""en"", use_cache=True\n)\n\nrom transformers import Seq2SeqTrainingArguments\n\ntraining_args = Seq2SeqTrainingArguments(\n#training_args = TrainingArguments(\n #Diag\n output_dir=""./whisper-medium-tp-test"", # name on the HF Hub\n per_device_train_batch_size=16,\n gradient_accumulation_steps=8, # increase by 2x for every 2x decrease in batch size\n learning_rate=1e-5,\n lr_scheduler_type=""constant_with_warmup"",\n warmup_steps=50,\n #Diag\n max_steps=50, # increase to 4000 if you have your own GPU or a Colab paid plan\n gradient_checkpointing=True,\n fp16=False,\n fp16_full_eval=False,\n eval_strategy=""steps"",\n per_device_eval_batch_size=8,\n predict_with_generate=True,\n generation_max_length=225,\n #Diag\n save_steps=50,\n eval_steps=10,\n logging_steps=10,\n report_to=[""tensorboard""],\n save_strategy=""steps"",\n #Diag\n load_best_model_at_end=False,\n metric_for_best_model=""wer"",\n greater_is_better=False,\n #Diag\n push_to_hub=False,\n)\n\nfrom transformers import Seq2SeqTrainer\n\n#Diag\nsmall_train_dataset = dataset[""train""].select(range(10)) # Select first 10 samples\nsmall_eval_dataset = dataset[""test""].select(range(10)) # Select first 10 samples\n\n\ntrainer = Seq2SeqTrainer(\n args=training_args,\n model=model,\n #Diag\n train_dataset=small_train_dataset,\n eval_dataset=small_eval_dataset,\n data_collator=data_collator,\n compute_metrics=compute_metrics,\n tokenizer=processor,\n)\ntrainer.train()\n#trainer.push_to_hub()\nSeems KV cache conflicts with gradient checkpointing graphs…
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-22T00:39:31.616Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 168571, 'topic_slug': 'runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/whisper_oom_kv.md', 'internal': False, 'reflection': False, 'title': 'whisper_oom_kv.md · John6666/forum1 at main', 'clicks': 8}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true/168571/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242375, 'name': 'Brian', 'username': 'brianko', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/3da27b/{size}.png', 'created_at': '2025-09-22T01:47:58.800Z', 'cooked': 'Wow, appreciate you putting all together in one place. I see several things I need to modify, will report back with success or failure (hopefully the former).
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-22T01:47:58.800Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168571, 'topic_slug': 'runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true', 'display_username': 'Brian', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true/168571/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242379, 'name': 'Brian', 'username': 'brianko', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/3da27b/{size}.png', 'created_at': '2025-09-22T04:39:28.523Z', 'cooked': 'Success!
\nThe significant changes I made based on your example were:
\ngradient_checkpointing_kwargs={""use_reentrant"": False}, \nfp16=False, \nfp16_full_eval=False,\nand I removed the model_generate = partial(…) call. That resolved the issue. Thank you!
Should I go ahead and try your other suggestions as well? I’m so pumped that it’s running that I don’t want to break it again…
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-22T04:40:35.083Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 20.4, 'yours': False, 'topic_id': 168571, 'topic_slug': 'runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true', 'display_username': 'Brian', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true/168571/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 242380, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-22T08:02:17.619Z', 'cooked': 'I think it’s best to copy stable code somewhere first before making changes. That’s what I always do. It gets messy though…
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-22T08:02:17.619Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168571, 'topic_slug': 'runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true/168571/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242399, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-22T20:02:56.971Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-09-22T20:02:56.971Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168571, 'topic_slug': 'runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/runtimeerror-backward-through-graph-with-whisper-medium-and-gradient-checkpointing-true/168571/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to fine-tune Whisper-medium and am getting this specific error during trainer.train():
tmp/ipython-input-774985985.py:8: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `Seq2SeqTrainer.__init__`. Use `processing_class` instead.
+ trainer = Seq2SeqTrainer(
+---------------------------------------------------------------------------
+RuntimeError Traceback (most recent call last)
+/tmp/ipython-input-774985985.py in <cell line: 0>()
+ 16 tokenizer=processor,
+ 17 )
+---> 18 trainer.train()
+ 19 #trainer.push_to_hub()
+
+10 frames
+/usr/local/lib/python3.12/dist-packages/torch/autograd/graph.py in _engine_run_backward(t_outputs, *args, **kwargs)
+ 827 unregister_hooks = _register_logging_hooks_on_whole_graph(t_outputs)
+ 828 try:
+--> 829 return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
+ 830 t_outputs, *args, **kwargs
+ 831 ) # Calls into the C++ engine to run the backward pass
+
+RuntimeError: Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). Saved intermediate values of the graph are freed when you call .backward() or autograd.grad(). Specify retain_graph=True if you need to backward through the graph a second time or if you need to access saved tensors after calling backward.
+These are the steps I’ve tried:
+Gradient checkpointing enabled (gradient_checkpointing=True).
FP16 disabled (fp16=False).
use_cache=False (which is the default for training with checkpointing, but you can mention you checked).
predict_with_generate=True.
Running on a minimal dataset subset.
+Using the original openai/whisper-medium model.
Restarting the runtime.
+Env:
+PyTorch version: 2.8.0+cu126
+Transformers version: 4.56.2
+Accelerate version: 1.10.1
+Datasets version: 4.1.1
+Modified code (per Gemini):
+from transformers import WhisperForConditionalGeneration
+# Diag
+from accelerate import Accelerator
+accelerator = Accelerator()
+device = accelerator.device
+
+model = WhisperForConditionalGeneration.from_pretrained(""openai/whisper-medium"")
+
+#Diag
+model.to(device)
+
+from functools import partial
+
+# disable cache during training since it's incompatible with gradient checkpointing
+model.config.use_cache = False
+
+# set language and task for generation and re-enable cache
+model.generate = partial(
+ model.generate, language=""en"", use_cache=True
+)
+
+rom transformers import Seq2SeqTrainingArguments
+
+training_args = Seq2SeqTrainingArguments(
+#training_args = TrainingArguments(
+ #Diag
+ output_dir=""./whisper-medium-tp-test"", # name on the HF Hub
+ per_device_train_batch_size=16,
+ gradient_accumulation_steps=8, # increase by 2x for every 2x decrease in batch size
+ learning_rate=1e-5,
+ lr_scheduler_type=""constant_with_warmup"",
+ warmup_steps=50,
+ #Diag
+ max_steps=50, # increase to 4000 if you have your own GPU or a Colab paid plan
+ gradient_checkpointing=True,
+ fp16=False,
+ fp16_full_eval=False,
+ eval_strategy=""steps"",
+ per_device_eval_batch_size=8,
+ predict_with_generate=True,
+ generation_max_length=225,
+ #Diag
+ save_steps=50,
+ eval_steps=10,
+ logging_steps=10,
+ report_to=[""tensorboard""],
+ save_strategy=""steps"",
+ #Diag
+ load_best_model_at_end=False,
+ metric_for_best_model=""wer"",
+ greater_is_better=False,
+ #Diag
+ push_to_hub=False,
+)
+
+from transformers import Seq2SeqTrainer
+
+#Diag
+small_train_dataset = dataset[""train""].select(range(10)) # Select first 10 samples
+small_eval_dataset = dataset[""test""].select(range(10)) # Select first 10 samples
+
+
+trainer = Seq2SeqTrainer(
+ args=training_args,
+ model=model,
+ #Diag
+ train_dataset=small_train_dataset,
+ eval_dataset=small_eval_dataset,
+ data_collator=data_collator,
+ compute_metrics=compute_metrics,
+ tokenizer=processor,
+)
+trainer.train()
+#trainer.push_to_hub()
+Success!
+The significant changes I made based on your example were:
+gradient_checkpointing_kwargs={""use_reentrant"": False},
+fp16=False,
+fp16_full_eval=False,
+and I removed the model_generate = partial(…) call. That resolved the issue. Thank you!
Should I go ahead and try your other suggestions as well? I’m so pumped that it’s running that I don’t want to break it again…
" +Fail to push README.md updates in Hugging Face Spaces,https://discuss.huggingface.co/t/fail-to-push-readme-md-updates-in-hugging-face-spaces/37992,37992,24,2023-04-28 06:30:45.291000+00:00,"[{'id': 66957, 'name': 'Hyoung-Kyu Song', 'username': 'deepkyu', 'avatar_template': '/user_avatar/discuss.huggingface.co/deepkyu/{size}/19615_2.png', 'created_at': '2023-04-28T06:30:45.351Z', 'cooked': 'Hi,
\nI tried to update a README.md file in my private Hugging Face Spaces.
\nBut I failed to push my commit which contains updating yaml card information with the following message:
remote: -------------------------------------------------------------------------\nremote: Unexpected internal error hook: yaml. (Request ID: 01GZ38NG4X5ER3VYAXBT65PC26)\nremote: -------------------------------------------------------------------------\nTo https://huggingface.co/spaces/nota-ai/efficient_wav2lip\n ! [remote rejected] main -> main (pre-receive hook declined)\nerror: failed to push some refs to \'https://huggingface.co/spaces/nota-ai/efficient_wav2lip\'\nAfter then, I came back to my browser and directly update with edit in Hugging Face Spaces.
\nLikewise, it shows an error without any message but a red “Error” box…
It seems that there are some issues in generating the space card with the front matter (yaml format at the top of README file).
\nThanks in advance.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2023-04-28T06:30:45.351Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 227, 'reads': 25, 'readers_count': 24, 'score': 1130.0, 'yours': False, 'topic_id': 37992, 'topic_slug': 'fail-to-push-readme-md-updates-in-hugging-face-spaces', 'display_username': 'Hyoung-Kyu Song', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/b/b388d6ede3659cb85d55ed299a127000fcd9b18b.png', 'internal': False, 'reflection': False, 'title': 'b388d6ede3659cb85d55ed299a127000fcd9b18b.png', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8000, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fail-to-push-readme-md-updates-in-hugging-face-spaces/37992/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 67034, 'name': 'Hyoung-Kyu Song', 'username': 'deepkyu', 'avatar_template': '/user_avatar/discuss.huggingface.co/deepkyu/{size}/19615_2.png', 'created_at': '2023-04-28T13:45:14.896Z', 'cooked': 'I tried it again and now it works.
\nI’ll close this issue.
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2023-04-28T13:45:14.896Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 21, 'readers_count': 20, 'score': 34.2, 'yours': False, 'topic_id': 37992, 'topic_slug': 'fail-to-push-readme-md-updates-in-hugging-face-spaces', 'display_username': 'Hyoung-Kyu Song', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8000, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fail-to-push-readme-md-updates-in-hugging-face-spaces/37992/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 67080, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-04-28T18:30:59.689Z', 'cooked': 'sorry we had an internal DNS issue
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2023-04-28T18:30:59.689Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 22, 'readers_count': 21, 'score': 34.4, 'yours': False, 'topic_id': 37992, 'topic_slug': 'fail-to-push-readme-md-updates-in-hugging-face-spaces', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 8000, 'username': 'deepkyu', 'name': 'Hyoung-Kyu Song', 'avatar_template': '/user_avatar/discuss.huggingface.co/deepkyu/{size}/19615_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fail-to-push-readme-md-updates-in-hugging-face-spaces/37992/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242290, 'name': 'Arun Baigra', 'username': 'arunbaigra', 'avatar_template': '/user_avatar/discuss.huggingface.co/arunbaigra/{size}/54048_2.png', 'created_at': '2025-09-19T11:42:13.201Z', 'cooked': 'help im facing the same error , pushed my files to the hf spaces but its showing configuration error i dont understand , help!
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-19T11:42:13.201Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 37992, 'topic_slug': 'fail-to-push-readme-md-updates-in-hugging-face-spaces', 'display_username': 'Arun Baigra', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 104117, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fail-to-push-readme-md-updates-in-hugging-face-spaces/37992/4', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242291, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-19T11:47:11.891Z', 'cooked': 'what error message?
', 'post_number': 5, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-19T11:47:11.891Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 37992, 'topic_slug': 'fail-to-push-readme-md-updates-in-hugging-face-spaces', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fail-to-push-readme-md-updates-in-hugging-face-spaces/37992/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I tried to update a README.md file in my private Hugging Face Spaces.
+But I failed to push my commit which contains updating yaml card information with the following message:
remote: -------------------------------------------------------------------------
+remote: Unexpected internal error hook: yaml. (Request ID: 01GZ38NG4X5ER3VYAXBT65PC26)
+remote: -------------------------------------------------------------------------
+To https://huggingface.co/spaces/nota-ai/efficient_wav2lip
+ ! [remote rejected] main -> main (pre-receive hook declined)
+error: failed to push some refs to 'https://huggingface.co/spaces/nota-ai/efficient_wav2lip'
+After then, I came back to my browser and directly update with edit in Hugging Face Spaces.
+Likewise, it shows an error without any message but a red “Error” box…
It seems that there are some issues in generating the space card with the front matter (yaml format at the top of README file).
+Thanks in advance.
","I tried it again and now it works.
+I’ll close this issue.
" +The best model is not being saved,https://discuss.huggingface.co/t/the-best-model-is-not-being-saved/168528,168528,5,2025-09-18 14:00:56.645000+00:00,"[{'id': 242243, 'name': 'Alex', 'username': 'SuperBowser', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9f8e36/{size}.png', 'created_at': '2025-09-18T14:00:56.730Z', 'cooked': 'I am using custom metric and in my training arguments I have
\ngreater_is_better=True,\nload_best_model_at_end=True,\nBut as far as I can the best model is not being saved. Here is link to my Colab notebook:
\n\nAnd here are all the details just in case:
\nMy platform and system data:
\nplatform: Linux
\nrelease: 6.1.123+
\nversion: #1 SMP PREEMPT_DYNAMIC Sun Mar 30 16:01:29 UTC 2025
\nmachine: x86_64
\ntorch: 2.8.0+cu126
\ntransformers:4.55.4
\ncompiler: 3.12.11 (main, Jun 4 2025, 08:56:18) [GCC 11.4.0]
\nGPU/TPU: Tesla T4
\nCUDA compiler:
\nnvcc: NVIDIA (R) Cuda compiler driver
\nCopyright (c) 2005-2024 NVIDIA Corporation
\nBuilt on Thu_Jun__6_02:18:23_PDT_2024
\nCuda compilation tools, release 12.5, V12.5.82
\nBuild cuda_12.5.r12.5/compiler.34385749_0
Here is my code:
\nfrom transformers import AutoModelForSequenceClassification, AutoTokenizer\nimport transformersimport sysimport torch\nimport pandas as pd, numpy as npfrom sklearn.preprocessing\nimport LabelEncoder\nimport joblibimport pandas as pd\nimport os\nfrom sklearn.model_selection import train_test_split\nfrom datasets import Datasetimport numpy as np\nfrom transformers import TrainingArguments,Trainer\nimport platform\n\nimport os\nmodel_name = \'microsoft/deberta-v3-xsmall\'\nmodel_name_path = \'deberta-v3-xsmall\'\nDIR = \'../MAP_models/\'+model_name_path+\'/tuned/\'\nos.makedirs(\'../MAP_models\', exist_ok = True)\nos.makedirs(\'../MAP_models/\'+model_name_path, exist_ok = True)\nos.makedirs(\'../MAP_models/\'+model_name_path+\'/tuned\', exist_ok=True)\nos.makedirs(\'../MAP_models/\'+model_name_path+\'/tuned/model\', exist_ok=True)\n\n\nNUM_LABELS = 65\ntext = [f""example {i}"" for i in range(300)]\nlabel = [i % NUM_LABELS for i in range(300)]\ntrain = pd.DataFrame({\'text\': text, \'label\': label})\n\ntrain_df, val_df = train_test_split(train, test_size=0.2, random_state=42)\n\n# Convert to Hugging Face Dataset\nCOLS = [\'text\',\'label\']\ntrain_ds = Dataset.from_pandas(train_df[COLS])\nval_ds = Dataset.from_pandas(val_df[COLS])\n\n\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nMAX_LEN = 256\n \n# Tokenization function\ndef tokenize(batch):\n return tokenizer(batch[""text""], padding=""max_length"", truncation=True, max_length=256)\n \ntrain_ds = train_ds.map(tokenize, batched=True)\nval_ds = val_ds.map(tokenize, batched=True)\n \n# Set format for PyTorch\ncolumns = [\'input_ids\', \'attention_mask\', \'label\']\ntrain_ds.set_format(type=\'torch\', columns=columns)\nval_ds.set_format(type=\'torch\', columns=columns)\n\nmodel = AutoModelForSequenceClassification.from_pretrained(\n model_name,\n num_labels=NUM_LABELS, trust_remote_code=True\n )\n\ndef compute_map3(eval_pred):\n logits, labels = eval_pred\n probs = torch.nn.functional.softmax(torch.tensor(logits), dim=-1).numpy()\n \n top3 = np.argsort(-probs, axis=1)[:, :3] # Top 3 predictions\n match = (top3 == labels[:, None])\n\n # Compute MAP@3 manually\n map3 = 0\n for i in range(len(labels)):\n if match[i, 0]:\n map3 += 1.0\n elif match[i, 1]:\n map3 += 1.0 / 2\n elif match[i, 2]:\n map3 += 1.0 / 3\n return {""map@3"": map3 / len(labels)}\n\nargs = TrainingArguments(\n per_device_train_batch_size = 2, \n per_device_eval_batch_size= 2,\n gradient_accumulation_steps = 1,\n warmup_steps = 10,\n num_train_epochs = 1,\n learning_rate = 5e-5,\n fp16 = True,\n bf16 = False,\n logging_steps = 1,\n optim = ""adamw_torch_fused"",\n weight_decay = 0.01,\n eval_strategy=""steps"",\n lr_scheduler_type = ""cosine_with_restarts"",\n seed = 3407,\n output_dir = DIR+""output"",\n logging_dir=DIR+""logs"",\n greater_is_better=True,\n load_best_model_at_end=True,\n save_steps=10,\n eval_steps=10,\n save_total_limit=3,\n report_to = ""none"", \n )\n\ntrainer = Trainer(\n model = model,\n processing_class = tokenizer,\n eval_dataset = val_ds,\n train_dataset = train_ds,\n args = args,\n compute_metrics = compute_map3,\n)\n\ntrainer_stats = trainer.train()\n\n\nIt produces the following output
\nStep\tTraining Loss\tValidation Loss\tMap@3
\n10\t4.235900\t4.182212\t0.025000
\n20\t4.245500\t4.176703\t0.038889
\n30\t4.166400\t4.171503\t0.030556
\n40\t4.163400\t4.174795\t0.025000
\n50\t4.187000\t4.174973\t0.025000
\n60\t4.240600\t4.176061\t0.038889
\n70\t4.123800\t4.177481\t0.036111
\n80\t4.130100\t4.177088\t0.033333
\n90\t4.140700\t4.177318\t0.022222
\n100\t4.180000\t4.178491\t0.022222
\n110\t4.112100\t4.178146\t0.025000
\n120\t4.229100\t4.178137\t0.025000
But when I run
\ntrainer.evaluate(val_ds)
{‘eval_loss’: 4.1822123527526855,
\n‘eval_map@3’: 0.025,
\n‘eval_runtime’: 0.9703,
\n‘eval_samples_per_second’: 61.836,
\n‘eval_steps_per_second’: 30.918,
\n‘epoch’: 1.0}
It seems like evaluation is done on the very first 10 steps, rather than on the best model.
\nWhat am I doing wrong?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-18T14:02:06.119Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 9, 'readers_count': 8, 'score': 36.8, 'yours': False, 'topic_id': 168528, 'topic_slug': 'the-best-model-is-not-being-saved', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://colab.research.google.com/drive/1ehTt53xlGV0Byx6yelifdEZcSgFREncy?usp=drive_link', 'internal': False, 'reflection': False, 'title': 'Google Colab', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-best-model-is-not-being-saved/168528/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242254, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-18T15:10:23.889Z', 'cooked': 'Due to metric_for_best_model is missing, etc. ?
Thank you so much! What a blunder!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-18T15:30:32.007Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 168528, 'topic_slug': 'the-best-model-is-not-being-saved', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-best-model-is-not-being-saved/168528/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242284, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-19T03:31:12.250Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-09-19T03:31:12.250Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168528, 'topic_slug': 'the-best-model-is-not-being-saved', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/the-best-model-is-not-being-saved/168528/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am using custom metric and in my training arguments I have
+greater_is_better=True,
+load_best_model_at_end=True,
+But as far as I can the best model is not being saved. Here is link to my Colab notebook:
+ +And here are all the details just in case:
+My platform and system data:
+platform: Linux
+release: 6.1.123+
+version: #1 SMP PREEMPT_DYNAMIC Sun Mar 30 16:01:29 UTC 2025
+machine: x86_64
+torch: 2.8.0+cu126
+transformers:4.55.4
+compiler: 3.12.11 (main, Jun 4 2025, 08:56:18) [GCC 11.4.0]
+GPU/TPU: Tesla T4
+CUDA compiler:
+nvcc: NVIDIA (R) Cuda compiler driver
+Copyright (c) 2005-2024 NVIDIA Corporation
+Built on Thu_Jun__6_02:18:23_PDT_2024
+Cuda compilation tools, release 12.5, V12.5.82
+Build cuda_12.5.r12.5/compiler.34385749_0
Here is my code:
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
+import transformersimport sysimport torch
+import pandas as pd, numpy as npfrom sklearn.preprocessing
+import LabelEncoder
+import joblibimport pandas as pd
+import os
+from sklearn.model_selection import train_test_split
+from datasets import Datasetimport numpy as np
+from transformers import TrainingArguments,Trainer
+import platform
+
+import os
+model_name = 'microsoft/deberta-v3-xsmall'
+model_name_path = 'deberta-v3-xsmall'
+DIR = '../MAP_models/'+model_name_path+'/tuned/'
+os.makedirs('../MAP_models', exist_ok = True)
+os.makedirs('../MAP_models/'+model_name_path, exist_ok = True)
+os.makedirs('../MAP_models/'+model_name_path+'/tuned', exist_ok=True)
+os.makedirs('../MAP_models/'+model_name_path+'/tuned/model', exist_ok=True)
+
+
+NUM_LABELS = 65
+text = [f""example {i}"" for i in range(300)]
+label = [i % NUM_LABELS for i in range(300)]
+train = pd.DataFrame({'text': text, 'label': label})
+
+train_df, val_df = train_test_split(train, test_size=0.2, random_state=42)
+
+# Convert to Hugging Face Dataset
+COLS = ['text','label']
+train_ds = Dataset.from_pandas(train_df[COLS])
+val_ds = Dataset.from_pandas(val_df[COLS])
+
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+MAX_LEN = 256
+
+# Tokenization function
+def tokenize(batch):
+ return tokenizer(batch[""text""], padding=""max_length"", truncation=True, max_length=256)
+
+train_ds = train_ds.map(tokenize, batched=True)
+val_ds = val_ds.map(tokenize, batched=True)
+
+# Set format for PyTorch
+columns = ['input_ids', 'attention_mask', 'label']
+train_ds.set_format(type='torch', columns=columns)
+val_ds.set_format(type='torch', columns=columns)
+
+model = AutoModelForSequenceClassification.from_pretrained(
+ model_name,
+ num_labels=NUM_LABELS, trust_remote_code=True
+ )
+
+def compute_map3(eval_pred):
+ logits, labels = eval_pred
+ probs = torch.nn.functional.softmax(torch.tensor(logits), dim=-1).numpy()
+
+ top3 = np.argsort(-probs, axis=1)[:, :3] # Top 3 predictions
+ match = (top3 == labels[:, None])
+
+ # Compute MAP@3 manually
+ map3 = 0
+ for i in range(len(labels)):
+ if match[i, 0]:
+ map3 += 1.0
+ elif match[i, 1]:
+ map3 += 1.0 / 2
+ elif match[i, 2]:
+ map3 += 1.0 / 3
+ return {""map@3"": map3 / len(labels)}
+
+args = TrainingArguments(
+ per_device_train_batch_size = 2,
+ per_device_eval_batch_size= 2,
+ gradient_accumulation_steps = 1,
+ warmup_steps = 10,
+ num_train_epochs = 1,
+ learning_rate = 5e-5,
+ fp16 = True,
+ bf16 = False,
+ logging_steps = 1,
+ optim = ""adamw_torch_fused"",
+ weight_decay = 0.01,
+ eval_strategy=""steps"",
+ lr_scheduler_type = ""cosine_with_restarts"",
+ seed = 3407,
+ output_dir = DIR+""output"",
+ logging_dir=DIR+""logs"",
+ greater_is_better=True,
+ load_best_model_at_end=True,
+ save_steps=10,
+ eval_steps=10,
+ save_total_limit=3,
+ report_to = ""none"",
+ )
+
+trainer = Trainer(
+ model = model,
+ processing_class = tokenizer,
+ eval_dataset = val_ds,
+ train_dataset = train_ds,
+ args = args,
+ compute_metrics = compute_map3,
+)
+
+trainer_stats = trainer.train()
+
+
+It produces the following output
+Step Training Loss Validation Loss Map@3
+10 4.235900 4.182212 0.025000
+20 4.245500 4.176703 0.038889
+30 4.166400 4.171503 0.030556
+40 4.163400 4.174795 0.025000
+50 4.187000 4.174973 0.025000
+60 4.240600 4.176061 0.038889
+70 4.123800 4.177481 0.036111
+80 4.130100 4.177088 0.033333
+90 4.140700 4.177318 0.022222
+100 4.180000 4.178491 0.022222
+110 4.112100 4.178146 0.025000
+120 4.229100 4.178137 0.025000
But when I run
+trainer.evaluate(val_ds)
{‘eval_loss’: 4.1822123527526855,
+‘eval_map@3’: 0.025,
+‘eval_runtime’: 0.9703,
+‘eval_samples_per_second’: 61.836,
+‘eval_steps_per_second’: 30.918,
+‘epoch’: 1.0}
It seems like evaluation is done on the very first 10 steps, rather than on the best model.
+What am I doing wrong?
","Due to metric_for_best_model is missing, etc. ?
Hello Everyone. I am a beginner learning LLMs and got hold of Book by Jay Alammar. I am trying to replicate the code in Colab, given by the author in the first chapter but I am not able to make it work. Looks like the latest version of transformers module had removed some functions and methods. It’s a simple code.
\n```\n# Check the version of the transformers library\nimport transformers\nprint(""Transformers version:"", transformers.__version__)\n# output in Colab shows \'Transformers version: 4.56.1\'\n\n# It\'s also good practice to check torch (PyTorch) version\nimport torch\nprint(""PyTorch version:"", torch.__version__)\n# output in Colab shows \'PyTorch version: 2.8.0+cu126\'\n\nfrom transformers import AutoModelForCausalLM, AutoTokenizer, pipeline\n\n#Load Model & Tokenizer\nmodel = AutoModelForCausalLM.from_pretrained(\n ""microsoft/Phi-3-mini-4k-instruct"",\n device_map = ""auto"",\n torch_dtype = ""auto"",\n trust_remote_code = True,\n)\n\ntokenizer = AutoTokenizer.from_pretrained(""microsoft/Phi-3-mini-4k-instruct"")\n\n#Create a pipeline\ngenerator = pipeline(\n ""text-generation"",\n model = model,\n tokenizer = tokenizer,\n return_full_text = False,\n max_new_tokens = 500,\n do_sample = False\n)\n\n# The prompt (user input/query)\nmessages = [\n {""role"": ""user"", ""content"": ""Create a funny joke about chickens.""}\n]\n\n# Generate Output\noutput = generator(messages)\nprint(output[0][\'generated_text\'])\n```\nHowever, the above code gives me the following error:
\n---------------------------------------------------------------------------\nAttributeError Traceback (most recent call last)\n/tmp/ipython-input-262462900.py in <cell line: 0>()\n 5 \n 6 # Generate Output\n----> 7 output = generator(messages)\n 8 print(output[0][\'generated_text\'])\n\n8 frames\n~/.cache/huggingface/modules/transformers_modules/microsoft/Phi-3-mini-4k-instruct/0a67737cc96d2554230f90338b163bc6380a2a85/modeling_phi3.py in prepare_inputs_for_generation(self, input_ids, past_key_values, attention_mask, inputs_embeds, **kwargs)\n 1289 if isinstance(past_key_values, Cache):\n 1290 cache_length = past_key_values.get_seq_length()\n-> 1291 past_length = past_key_values.seen_tokens\n 1292 max_cache_length = past_key_values.get_max_length()\n 1293 else:\n\nAttributeError: \'DynamicCache\' object has no attribute \'seen_tokens\'\nI tried modifying the code using ChatGPT, deepseek and inbuilt gemini as well, but they weren’t able to solve the problem. One of the solution they presented was to fall back on the transformer version (to 4.36.0), which i believe will not help me in the long term.
\nWhat could be the possible solution for this? Is the book really outdated after its release 11 months ago? Please Help! I’m not able to proceed further.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-15T11:16:06.575Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 132, 'reads': 5, 'readers_count': 4, 'score': 591.0, 'yours': False, 'topic_id': 168439, 'topic_slug': 'cannot-solve-dynamiccache-seen-tokens-error', 'display_username': 'Zarem Nacim', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103825, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-solve-dynamiccache-seen-tokens-error/168439/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 242014, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-15T12:17:44.040Z', 'cooked': 'Downgrading is fine, but if you want to run it on the latest Transformers, this method might be better. Since PHI-3 should be supported by default now, I don’t think remote_code is necessary for this model anymore…
model = AutoModelForCausalLM.from_pretrained(\n ""microsoft/Phi-3-mini-4k-instruct"",\n device_map = ""auto"",\n torch_dtype = ""auto"",\n # trust_remote_code = True, <= delete this line to avoid using outdated code\n)\nThanks a lot. You saved my day. I was having a tough time figuring this out. BTW, what could be the problem with this line of code?
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-15T15:31:11.417Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 168439, 'topic_slug': 'cannot-solve-dynamiccache-seen-tokens-error', 'display_username': 'Zarem Nacim', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103825, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-solve-dynamiccache-seen-tokens-error/168439/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 242044, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-15T21:28:48.986Z', 'cooked': '\n\nwhat could be the problem with this line of code?
\n
Setting trust_remote_code=True causes the class from the .py file in the Hugging Face model repo to be used, so if that code is outdated, the old code will be used.
It’s useful for new models that aren’t officially supported or for customized models, but it’s unnecessary if the current version provides support in default.
Usually, code rarely becomes unusable due to Transoformers version upgrades, but around version 4.49.0 there was a major refactoring, so function locations changed and errors can occur. I occasionally pin the version myself. pip install transformers<=4.48.3
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-09-16T09:29:38.566Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168439, 'topic_slug': 'cannot-solve-dynamiccache-seen-tokens-error', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cannot-solve-dynamiccache-seen-tokens-error/168439/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello Everyone. I am a beginner learning LLMs and got hold of Book by Jay Alammar. I am trying to replicate the code in Colab, given by the author in the first chapter but I am not able to make it work. Looks like the latest version of transformers module had removed some functions and methods. It’s a simple code.
+```
+# Check the version of the transformers library
+import transformers
+print(""Transformers version:"", transformers.__version__)
+# output in Colab shows 'Transformers version: 4.56.1'
+
+# It's also good practice to check torch (PyTorch) version
+import torch
+print(""PyTorch version:"", torch.__version__)
+# output in Colab shows 'PyTorch version: 2.8.0+cu126'
+
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
+
+#Load Model & Tokenizer
+model = AutoModelForCausalLM.from_pretrained(
+ ""microsoft/Phi-3-mini-4k-instruct"",
+ device_map = ""auto"",
+ torch_dtype = ""auto"",
+ trust_remote_code = True,
+)
+
+tokenizer = AutoTokenizer.from_pretrained(""microsoft/Phi-3-mini-4k-instruct"")
+
+#Create a pipeline
+generator = pipeline(
+ ""text-generation"",
+ model = model,
+ tokenizer = tokenizer,
+ return_full_text = False,
+ max_new_tokens = 500,
+ do_sample = False
+)
+
+# The prompt (user input/query)
+messages = [
+ {""role"": ""user"", ""content"": ""Create a funny joke about chickens.""}
+]
+
+# Generate Output
+output = generator(messages)
+print(output[0]['generated_text'])
+```
+However, the above code gives me the following error:
+---------------------------------------------------------------------------
+AttributeError Traceback (most recent call last)
+/tmp/ipython-input-262462900.py in <cell line: 0>()
+ 5
+ 6 # Generate Output
+----> 7 output = generator(messages)
+ 8 print(output[0]['generated_text'])
+
+8 frames
+~/.cache/huggingface/modules/transformers_modules/microsoft/Phi-3-mini-4k-instruct/0a67737cc96d2554230f90338b163bc6380a2a85/modeling_phi3.py in prepare_inputs_for_generation(self, input_ids, past_key_values, attention_mask, inputs_embeds, **kwargs)
+ 1289 if isinstance(past_key_values, Cache):
+ 1290 cache_length = past_key_values.get_seq_length()
+-> 1291 past_length = past_key_values.seen_tokens
+ 1292 max_cache_length = past_key_values.get_max_length()
+ 1293 else:
+
+AttributeError: 'DynamicCache' object has no attribute 'seen_tokens'
+I tried modifying the code using ChatGPT, deepseek and inbuilt gemini as well, but they weren’t able to solve the problem. One of the solution they presented was to fall back on the transformer version (to 4.36.0), which i believe will not help me in the long term.
+What could be the possible solution for this? Is the book really outdated after its release 11 months ago? Please Help! I’m not able to proceed further.
","Downgrading is fine, but if you want to run it on the latest Transformers, this method might be better. Since PHI-3 should be supported by default now, I don’t think remote_code is necessary for this model anymore…
model = AutoModelForCausalLM.from_pretrained(
+ ""microsoft/Phi-3-mini-4k-instruct"",
+ device_map = ""auto"",
+ torch_dtype = ""auto"",
+ # trust_remote_code = True, <= delete this line to avoid using outdated code
+)
+What’s the definiation of lazy loading? Do the IterableDataset and Dataset decided whether there is the lazy loading? I think lazy loading is that we don’t load all the data at the same time. So only we used IterableDataset , lazy loading will happen.
\nAnother question comes out. Does IterableDataset use memory-mapping and zero-copy to retrive data? Will both IterableDataset and Dataset occupy the same RAM when loading the same datasets? If we just retrive data without shuffle and locally, the speed differece between IterableDataset and Dataset is because contiguous sequential access is faster than random access, right?
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-12T14:13:23.944Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 6, 'readers_count': 5, 'score': 46.2, 'yours': False, 'topic_id': 168304, 'topic_slug': 'what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally', 'display_username': 'Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 59867, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally/168304/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241789, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-12T14:50:56.300Z', 'cooked': 'Aside from definitions and general aspects, I think only the author or maintainer can really understand the implementation… @lhoestq
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-12T14:50:56.300Z', 'reply_count': 2, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 31.2, 'yours': False, 'topic_id': 168304, 'topic_slug': 'what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/lazy_loading.md', 'internal': False, 'reflection': False, 'title': 'lazy_loading.md · John6666/forum1 at main', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally/168304/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241808, 'name': 'Zhao', 'username': 'Zoe0427', 'avatar_template': '/user_avatar/discuss.huggingface.co/zoe0427/{size}/53729_2.png', 'created_at': '2025-09-12T19:24:34.673Z', 'cooked': 'Thank you John! That link is very helpful!
\nThere is a confusion about: “But one caveat is that you must have the entire dataset stored on your disk or in memory, which blocks you from accessing datasets bigger than the disk.” Does memory refer to RAM? I can understand dataset is larger than disk, but I think load_dataset can covert other file format to .arrow, and it occupied low RAM, right?
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-12T19:24:34.673Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 6, 'readers_count': 5, 'score': 36.2, 'yours': False, 'topic_id': 168304, 'topic_slug': 'what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally', 'display_username': 'Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/datasets/en/about_mapstyle_vs_iterable', 'internal': False, 'reflection': False, 'title': 'Differences between Dataset and IterableDataset', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 59867, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally/168304/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241810, 'name': 'Zhao', 'username': 'Zoe0427', 'avatar_template': '/user_avatar/discuss.huggingface.co/zoe0427/{size}/53729_2.png', 'created_at': '2025-09-12T19:39:44.616Z', 'cooked': 'And also I noticed huge virtual memory(around 100G, and my dataset is also around 100G) is occupied when I use load_from_disk or load_dataset without streaming to load .arrow files. Is that normal? I see the blog, and for my understanding, zero_copy utilizes the virtual memory indeed, and the size of VM is related to the size of datasets, right?
\nThank you!
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-12T19:39:44.616Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 26.0, 'yours': False, 'topic_id': 168304, 'topic_slug': 'what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally', 'display_username': 'Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://cmmon.medium.com/the-zero-copy-frontier-a7d2a4e05127', 'internal': False, 'reflection': False, 'title': 'The Zero-Copy Frontier. When we hear the term Zero-copy, just… | by Aniket Kumar | Medium', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 59867, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally/168304/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241823, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-12T23:22:26.628Z', 'cooked': 'I’ve never worked with huge datasets…
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-12T23:22:26.628Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 168304, 'topic_slug': 'what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/lazy_loading2.md', 'internal': False, 'reflection': False, 'title': 'lazy_loading2.md · John6666/forum1 at main', 'clicks': 3}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally/168304/5', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 241848, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-13T11:22:53.141Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-09-13T11:22:53.141Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 168304, 'topic_slug': 'what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/what-s-the-definiation-of-lazy-loading-is-iterabledataset-also-faster-than-dataset-when-loading-locally/168304/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","What’s the definiation of lazy loading? Do the IterableDataset and Dataset decided whether there is the lazy loading? I think lazy loading is that we don’t load all the data at the same time. So only we used IterableDataset , lazy loading will happen.
+Another question comes out. Does IterableDataset use memory-mapping and zero-copy to retrive data? Will both IterableDataset and Dataset occupy the same RAM when loading the same datasets? If we just retrive data without shuffle and locally, the speed differece between IterableDataset and Dataset is because contiguous sequential access is faster than random access, right?
","I’ve never worked with huge datasets…
" +Getting started with Voxtral for ASR transcription,https://discuss.huggingface.co/t/getting-started-with-voxtral-for-asr-transcription/168281,168281,13,2025-09-11 03:33:04.077000+00:00,"[{'id': 241677, 'name': 'Georg Heiler', 'username': 'geoHeil', 'avatar_template': '/user_avatar/discuss.huggingface.co/geoheil/{size}/26801_2.png', 'created_at': '2025-09-11T03:33:04.141Z', 'cooked': 'I am trying to execute Voxtral the default example for transcription of the obama speech for ASR of Voxtral.
\nHow can this be changed so the real/full text is returned - not just the first word.
\nimport torch\nfrom transformers import VoxtralForConditionalGeneration, AutoProcessor, infer_device\n\ndevice = infer_device()\nrepo_id = ""mistralai/Voxtral-Mini-3B-2507""\n\nprocessor = AutoProcessor.from_pretrained(repo_id)\nmodel = VoxtralForConditionalGeneration.from_pretrained(repo_id, dtype=torch.bfloat16, device_map=device)\n\ninputs = processor.apply_transcription_request(language=""en"", audio=""https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/obama.mp3"", model_id=repo_id)\ninputs = inputs.to(device, dtype=torch.bfloat16)\n\noutputs = model.generate(**inputs, max_new_tokens=500)\ndecoded_outputs = processor.batch_decode(outputs[:, inputs.input_ids.shape[1]:], skip_special_tokens=True)\n\nprint(""\\nGenerated responses:"")\nprint(""="" * 80)\nfor decoded_output in decoded_outputs:\n print(decoded_output)\n print(""="" * 80)\n\n\nI think this is a bfloat 16 mixup with MPS
\nimport torch\nfrom transformers import VoxtralForConditionalGeneration, AutoProcessor\n\ndevice = ""mps"" if torch.backends.mps.is_available() else ""cpu""\nrepo_id = ""mistralai/Voxtral-Mini-3B-2507""\naudio_url = ""https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/obama.mp3""\n\nprocessor = AutoProcessor.from_pretrained(repo_id)\n\n# ⚠️ Use fp16 on MPS (avoid bf16). Also force eager attention on MPS for correctness.\nmodel = VoxtralForConditionalGeneration.from_pretrained(\n repo_id,\n torch_dtype=torch.float16 if device == ""mps"" else torch.float32,\n attn_implementation=""eager"", # helps avoid MPS SDPA quirks\n device_map={"""": device}, # single-device map; no auto-sharding on MPS\n)\n\n# Build the transcription request\ninputs = processor.apply_transcription_request(\n language=""en"", audio=audio_url, model_id=repo_id\n)\n\n# Move to device and cast only floating tensors to fp16 on MPS\ninputs = inputs.to(device) # move first\nfor k, v in list(inputs.items()):\n if torch.is_tensor(v) and torch.is_floating_point(v) and device == ""mps"":\n inputs[k] = v.to(dtype=torch.float16)\n\n# Greedy is fine for transcription; raise the budget for a ~5 min clip\noutputs = model.generate(**inputs, max_new_tokens=2048, do_sample=False)\n\ndecoded = processor.batch_decode(\n outputs[:, inputs.input_ids.shape[1]:],\n skip_special_tokens=True\n)\n\nprint(""\\nGenerated responses:\\n"" + ""=""*80)\nfor d in decoded:\n print(d)\n print(""=""*80)\n\nfixes things for me
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-09-11T03:46:54.017Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 2, 'readers_count': 1, 'score': 25.4, 'yours': False, 'topic_id': 168281, 'topic_slug': 'getting-started-with-voxtral-for-asr-transcription', 'display_username': 'Georg Heiler', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 49603, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/getting-started-with-voxtral-for-asr-transcription/168281/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 241714, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-11T15:47:30.722Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-09-11T15:47:30.722Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168281, 'topic_slug': 'getting-started-with-voxtral-for-asr-transcription', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/getting-started-with-voxtral-for-asr-transcription/168281/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to execute Voxtral the default example for transcription of the obama speech for ASR of Voxtral.
+How can this be changed so the real/full text is returned - not just the first word.
+import torch
+from transformers import VoxtralForConditionalGeneration, AutoProcessor, infer_device
+
+device = infer_device()
+repo_id = ""mistralai/Voxtral-Mini-3B-2507""
+
+processor = AutoProcessor.from_pretrained(repo_id)
+model = VoxtralForConditionalGeneration.from_pretrained(repo_id, dtype=torch.bfloat16, device_map=device)
+
+inputs = processor.apply_transcription_request(language=""en"", audio=""https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/obama.mp3"", model_id=repo_id)
+inputs = inputs.to(device, dtype=torch.bfloat16)
+
+outputs = model.generate(**inputs, max_new_tokens=500)
+decoded_outputs = processor.batch_decode(outputs[:, inputs.input_ids.shape[1]:], skip_special_tokens=True)
+
+print(""\nGenerated responses:"")
+print(""="" * 80)
+for decoded_output in decoded_outputs:
+ print(decoded_output)
+ print(""="" * 80)
+
+
+I think this is a bfloat 16 mixup with MPS
+import torch
+from transformers import VoxtralForConditionalGeneration, AutoProcessor
+
+device = ""mps"" if torch.backends.mps.is_available() else ""cpu""
+repo_id = ""mistralai/Voxtral-Mini-3B-2507""
+audio_url = ""https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/obama.mp3""
+
+processor = AutoProcessor.from_pretrained(repo_id)
+
+# ⚠️ Use fp16 on MPS (avoid bf16). Also force eager attention on MPS for correctness.
+model = VoxtralForConditionalGeneration.from_pretrained(
+ repo_id,
+ torch_dtype=torch.float16 if device == ""mps"" else torch.float32,
+ attn_implementation=""eager"", # helps avoid MPS SDPA quirks
+ device_map={"""": device}, # single-device map; no auto-sharding on MPS
+)
+
+# Build the transcription request
+inputs = processor.apply_transcription_request(
+ language=""en"", audio=audio_url, model_id=repo_id
+)
+
+# Move to device and cast only floating tensors to fp16 on MPS
+inputs = inputs.to(device) # move first
+for k, v in list(inputs.items()):
+ if torch.is_tensor(v) and torch.is_floating_point(v) and device == ""mps"":
+ inputs[k] = v.to(dtype=torch.float16)
+
+# Greedy is fine for transcription; raise the budget for a ~5 min clip
+outputs = model.generate(**inputs, max_new_tokens=2048, do_sample=False)
+
+decoded = processor.batch_decode(
+ outputs[:, inputs.input_ids.shape[1]:],
+ skip_special_tokens=True
+)
+
+print(""\nGenerated responses:\n"" + ""=""*80)
+for d in decoded:
+ print(d)
+ print(""=""*80)
+
+fixes things for me
" +Getting the Space name programmatically,https://discuss.huggingface.co/t/getting-the-space-name-programmatically/168253,168253,24,2025-09-10 09:20:15.719000+00:00,"[{'id': 241610, 'name': 'João Ricardo Silva', 'username': 'jrsilva', 'avatar_template': '/user_avatar/discuss.huggingface.co/jrsilva/{size}/53168_2.png', 'created_at': '2025-09-10T09:20:15.781Z', 'cooked': 'Is there a programmatic way of a Space knowing its own name?
\nFor instance, the restart_space method of the huggingface_hub API requires a repo_id. If, say, I want the Space to restart itself, is there a programmatic way of getting this repo_id (and thus working without requiring changes if the Space is ever renamed) or do I have to hard-code it?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-10T09:20:15.781Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 4, 'readers_count': 3, 'score': 65.8, 'yours': False, 'topic_id': 168253, 'topic_slug': 'getting-the-space-name-programmatically', 'display_username': 'João Ricardo Silva', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102714, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/getting-the-space-name-programmatically/168253/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241616, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-10T10:59:05.305Z', 'cooked': 'Maybe simply by this?
\nimport os\nspace_id = os.getenv(""SPACE_ID"", """") # e.g. ""username/space-name""\nYou are quite right. I somehow missed that part of the documentation. Thank you.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-10T12:04:43.563Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168253, 'topic_slug': 'getting-the-space-name-programmatically', 'display_username': 'João Ricardo Silva', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102714, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/getting-the-space-name-programmatically/168253/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241672, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-11T00:04:44.148Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-09-11T00:04:44.148Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168253, 'topic_slug': 'getting-the-space-name-programmatically', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/getting-the-space-name-programmatically/168253/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Is there a programmatic way of a Space knowing its own name?
+For instance, the restart_space method of the huggingface_hub API requires a repo_id. If, say, I want the Space to restart itself, is there a programmatic way of getting this repo_id (and thus working without requiring changes if the Space is ever renamed) or do I have to hard-code it?
","Maybe simply by this?
+import os
+space_id = os.getenv(""SPACE_ID"", """") # e.g. ""username/space-name""
+Hi I´m new here and new to transformers. I´m develloping app for information extraction from invoices using layoutlmv3 and I came to a problem. When I use layoutlmv3 processor to encode words from invoice and I pass the word_labels. The labels from the processor does not match the original dataset labels(before nor after removing -100 labels) but only in small parts…
\nExample:
\nI pass to encoder this word_labels: [0,0,0,1,0,0,3,4,0,5,0,0,0,0,11,0,0,0,13,0,0,15,0,0,17,…]
\nLabels from processor after encoding(removed -100): [0,0,0,1,0,0,3,4,0,5,0,0,0,0,11,0,0,0,0,13,0,0,15,0,0,17,…]
\nThe problem is that in original I have three zeroes between 11 and 13 and in the labels from processor I have four zeroes between 11 and 13. Do you someone, why is that happening? The rest of the labels is ok I think, but shifted because of that extra zero. Thanks for help or any advices.
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-09T09:43:15.399Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 10, 'reads': 2, 'readers_count': 1, 'score': 65.4, 'yours': False, 'topic_id': 168230, 'topic_slug': 'layoutlmv3-word-labels-does-not-match-original-labels-from-dataset', 'display_username': 'Tomáš', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103183, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/layoutlmv3-word-labels-does-not-match-original-labels-from-dataset/168230/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241551, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-09T12:52:48.041Z', 'cooked': 'Seems you’re comparing word-level labels to the processor’s token-level labels? Maybe.
\nfrom transformers import LayoutLMv3Processor\nfrom PIL import Image\n\n# --- toy invoice words, one value likely splits into multiple subwords ---\nwords = [""Invoice"", ""No."", ""12345"", ""Total"", ""USD"", ""1,234.56"", "".""]\nboxes = [\n [ 50, 50, 200, 100],\n [210, 50, 260, 100],\n [270, 50, 380, 100],\n [ 50, 150, 140, 200],\n [150, 150, 220, 200],\n [230, 150, 380, 200],\n [390, 150, 405, 200],\n]\n# 0 = O, 1 = INVOICE_NO, 3 = AMOUNT (example)\nword_labels = [0, 0, 1, 0, 0, 3, 0]\n\nimage = Image.new(""RGB"", (1000, 1000), ""white"")\nprocessor = LayoutLMv3Processor.from_pretrained(""microsoft/layoutlmv3-base"", apply_ocr=False)\n\n# ------------------\n# WRONG COMPARISON\n# ------------------\n# Make the tokenizer label *every* subword, so any split word duplicates its label.\nprocessor.tokenizer.only_label_first_subword = False\n\nenc_wrong = processor(\n images=image,\n text=words,\n boxes=boxes,\n word_labels=word_labels,\n truncation=True,\n padding=""max_length"",\n max_length=128,\n return_tensors=""pt"",\n)\n\nlabels_tok_wrong = enc_wrong[""labels""][0].tolist()\n# Naively drop -100 (special tokens, padding, or ignored subtokens)\nlabels_wrong_naive = [l for l in labels_tok_wrong if l != -100]\n\nprint(""WRONG: compare original vs processor labels after removing -100"")\nprint(""original:"", word_labels)\nprint(""encoded :"", labels_wrong_naive[:len(word_labels)+10]) # show a slice\nprint(""equal? "", word_labels == labels_wrong_naive)\n\n# ------------------\n# CORRECT COMPARISON (two valid options)\n# ------------------\n\n# Option A: Keep only first subword labels during encoding\nprocessor.tokenizer.only_label_first_subword = True\nenc_ok = processor(\n images=image,\n text=words,\n boxes=boxes,\n word_labels=word_labels,\n truncation=True,\n padding=""max_length"",\n max_length=128,\n return_tensors=""pt"",\n)\nlabels_tok_ok = enc_ok[""labels""][0].tolist()\nlabels_ok_naive = [l for l in labels_tok_ok if l != -100] # now this is 1:1 with words\nprint(""\\nCORRECT A: only_label_first_subword=True then drop -100"")\nprint(""original:"", word_labels)\nprint(""encoded :"", labels_ok_naive)\nprint(""equal? "", word_labels == labels_ok_naive)\n\n# Option B: Collapse token-level labels back to word-level using word_ids()\nword_ids = enc_wrong.word_ids(0) # from the earlier \'enc_wrong\' with duplicated subword labels\nrecovered = []\nseen = set()\nfor wid, lab in zip(word_ids, labels_tok_wrong):\n if wid is None or lab == -100:\n continue\n if wid not in seen: # first subword of each word only\n recovered.append(lab)\n seen.add(wid)\n\nprint(""\\nCORRECT B: collapse tokens -> words via word_ids() on any encoding"")\nprint(""original:"", word_labels)\nprint(""recovered:"", recovered)\nprint(""equal? "", word_labels == recovered)\n""""""\nWRONG: compare original vs processor labels after removing -100\noriginal: [0, 0, 1, 0, 0, 3, 0]\nencoded : [0, 0, 0, 0, 1, 1, 0, 0, 3, 3, 3, 3, 3, 0]\nequal? False\n\nCORRECT A: only_label_first_subword=True then drop -100\noriginal: [0, 0, 1, 0, 0, 3, 0]\nencoded : [0, 0, 1, 0, 0, 3, 0]\nequal? True\n\nCORRECT B: collapse tokens -> words via word_ids() on any encoding\noriginal: [0, 0, 1, 0, 0, 3, 0]\nrecovered: [0, 0, 1, 0, 0, 3, 0]\nequal? True\n""""""\nThank you for your answer, but I just few minutes back resolved my problem. Unfortunetly it was not caused by what you suggests. The problem was that the layoutlmv3 for some reason does not work well with dialects and I have my invoices in Czech, so it for example from word Plnění created three separate tokens: Pln ě ní and in my dataset I had only divided into Plně and ní. I´m not sure if my explanation is clear, but I don´t know how to say it otherwise. The solution was to use unidecode() on each word in my dataset before using processor.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-09T13:10:08.089Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 168230, 'topic_slug': 'layoutlmv3-word-labels-does-not-match-original-labels-from-dataset', 'display_username': 'Tomáš', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103183, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/layoutlmv3-word-labels-does-not-match-original-labels-from-dataset/168230/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 241600, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-10T01:10:22.869Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-09-10T01:10:22.869Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168230, 'topic_slug': 'layoutlmv3-word-labels-does-not-match-original-labels-from-dataset', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/layoutlmv3-word-labels-does-not-match-original-labels-from-dataset/168230/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi I´m new here and new to transformers. I´m develloping app for information extraction from invoices using layoutlmv3 and I came to a problem. When I use layoutlmv3 processor to encode words from invoice and I pass the word_labels. The labels from the processor does not match the original dataset labels(before nor after removing -100 labels) but only in small parts…
+Example:
+I pass to encoder this word_labels: [0,0,0,1,0,0,3,4,0,5,0,0,0,0,11,0,0,0,13,0,0,15,0,0,17,…]
+Labels from processor after encoding(removed -100): [0,0,0,1,0,0,3,4,0,5,0,0,0,0,11,0,0,0,0,13,0,0,15,0,0,17,…]
+The problem is that in original I have three zeroes between 11 and 13 and in the labels from processor I have four zeroes between 11 and 13. Do you someone, why is that happening? The rest of the labels is ok I think, but shifted because of that extra zero. Thanks for help or any advices.
","Thank you for your answer, but I just few minutes back resolved my problem. Unfortunetly it was not caused by what you suggests. The problem was that the layoutlmv3 for some reason does not work well with dialects and I have my invoices in Czech, so it for example from word Plnění created three separate tokens: Pln ě ní and in my dataset I had only divided into Plně and ní. I´m not sure if my explanation is clear, but I don´t know how to say it otherwise. The solution was to use unidecode() on each word in my dataset before using processor.
" +Image to text using blip2 gives incorrect answer,https://discuss.huggingface.co/t/image-to-text-using-blip2-gives-incorrect-answer/168177,168177,5,2025-09-07 15:31:05.250000+00:00,"[{'id': 241418, 'name': 'Raman Shah', 'username': 'rxshah', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/a587f6/{size}.png', 'created_at': '2025-09-07T15:31:05.323Z', 'cooked': 'Here is code snippet slightly modified from blip2 site:
\nfirst prompt “Question: How many cats are there? Answer:” –> gives correct answer Two
\nHowever, second prompt “Question: How many dogs are there? Answer:” –> gives incorrect answer - Two should be Zero or None.
\nIs this because the accuracy of the trained model is not 100% we should get incorrect answers? OR AM I doing something incorrectly?
\nHere is the complete code:
\nfrom PIL import Image
\nimport requests
\nfrom transformers import Blip2Processor, Blip2ForConditionalGeneration
\nimport torch
device = “cuda” if torch.cuda.is_available() else “cpu”
\nprocessor = Blip2Processor.from_pretrained(“Salesforce/blip2-opt-2.7b”)
\nmodel = Blip2ForConditionalGeneration.from_pretrained(
\n“Salesforce/blip2-opt-2.7b”, torch_dtype=torch.float16
\n)
\nmodel.to(device)
url = “http://images.cocodataset.org/val2017/000000039769.jpg”
\nimage = Image.open(requests.get(url, stream=True).raw)
prompt = “Question: How many cats are there? Answer:”
\ninputs = processor(images=image, text=prompt, return_tensors=“pt”).to(
\ndevice, torch.float16
\n)
outputs = model.generate(**inputs)
\ntext = processor.tokenizer.batch_decode(outputs, skip_special_tokens=True)
\nprint(text)
Gives correct answer: [‘Question: How many cats are there? Answer: Two\\n’]
\nHowever, when I change prompt to
\nprompt2 = ""Question: How many dogs are there? Answer: ""
\ninputs2 = processor(images=image, text=prompt2, return_tensors=“pt”).to(
\ndevice, torch.float16
\n)
outputs2 = model.generate(**inputs2)
\ntext2 = processor.tokenizer.batch_decode(outputs2, skip_special_tokens=True)
\nprint(text2)
[‘Question: How many dogs are there? Answer: Two\\n’]
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-07T15:45:45.288Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 6, 'readers_count': 5, 'score': 61.2, 'yours': False, 'topic_id': 168177, 'topic_slug': 'image-to-text-using-blip2-gives-incorrect-answer', 'display_username': 'Raman Shah', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://images.cocodataset.org/val2017/000000039769.jpg%E2%80%9D', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 80638, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/image-to-text-using-blip2-gives-incorrect-answer/168177/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241436, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-07T20:48:34.727Z', 'cooked': '\n\nOR AM I doing something incorrectly?
\n
There’s no problem with the code; it seems to be a known issue with the model / architecture. You might want to try using some fine-tuned version.
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-07T20:48:34.727Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 168177, 'topic_slug': 'image-to-text-using-blip2-gives-incorrect-answer', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/Salesforce/blip2-opt-2.7b-coco', 'internal': False, 'reflection': False, 'title': 'Salesforce/blip2-opt-2.7b-coco · Hugging Face', 'clicks': 2}, {'url': 'https://arxiv.org/pdf/2403.01373', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/image-to-text-using-blip2-gives-incorrect-answer/168177/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241443, 'name': 'Raman Shah', 'username': 'rxshah', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/a587f6/{size}.png', 'created_at': '2025-09-08T01:14:33.037Z', 'cooked': 'Thanks!!
\nTried the examples you pointed to. The number of dogs still gave Two. However, following the examples further got following results:
\n55.3% that image 0 is \'a photo of a cat\'\n44.7% that image 0 is \'a photo of a dog\'\nPerhaps this explains why the model cannot distinguish between cats, dogs or anything else?
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-08T01:14:33.037Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 168177, 'topic_slug': 'image-to-text-using-blip2-gives-incorrect-answer', 'display_username': 'Raman Shah', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 80638, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/image-to-text-using-blip2-gives-incorrect-answer/168177/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241446, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-08T03:51:52.414Z', 'cooked': 'Yeah. For example, CLIP can perfectly classify dogs and cats, but BLIP seems utterly unsuitable for classification…
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-08T03:51:52.414Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 168177, 'topic_slug': 'image-to-text-using-blip2-gives-incorrect-answer', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/John6666/forum1/blob/main/blip2_cats_dogs.md', 'internal': False, 'reflection': False, 'title': 'blip2_cats_dogs.md · John6666/forum1 at main', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/image-to-text-using-blip2-gives-incorrect-answer/168177/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 241472, 'name': 'Raman Shah', 'username': 'rxshah', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/a587f6/{size}.png', 'created_at': '2025-09-08T13:52:59.063Z', 'cooked': 'Thanks for the clear explanation!!
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-08T13:52:59.063Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168177, 'topic_slug': 'image-to-text-using-blip2-gives-incorrect-answer', 'display_username': 'Raman Shah', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 80638, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/image-to-text-using-blip2-gives-incorrect-answer/168177/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241501, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-09T01:53:46.094Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-09-09T01:53:46.094Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168177, 'topic_slug': 'image-to-text-using-blip2-gives-incorrect-answer', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/image-to-text-using-blip2-gives-incorrect-answer/168177/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Here is code snippet slightly modified from blip2 site:
+first prompt “Question: How many cats are there? Answer:” –> gives correct answer Two
+However, second prompt “Question: How many dogs are there? Answer:” –> gives incorrect answer - Two should be Zero or None.
+Is this because the accuracy of the trained model is not 100% we should get incorrect answers? OR AM I doing something incorrectly?
+Here is the complete code:
+from PIL import Image
+import requests
+from transformers import Blip2Processor, Blip2ForConditionalGeneration
+import torch
device = “cuda” if torch.cuda.is_available() else “cpu”
+processor = Blip2Processor.from_pretrained(“Salesforce/blip2-opt-2.7b”)
+model = Blip2ForConditionalGeneration.from_pretrained(
+“Salesforce/blip2-opt-2.7b”, torch_dtype=torch.float16
+)
+model.to(device)
url = “http://images.cocodataset.org/val2017/000000039769.jpg”
+image = Image.open(requests.get(url, stream=True).raw)
prompt = “Question: How many cats are there? Answer:”
+inputs = processor(images=image, text=prompt, return_tensors=“pt”).to(
+device, torch.float16
+)
outputs = model.generate(**inputs)
+text = processor.tokenizer.batch_decode(outputs, skip_special_tokens=True)
+print(text)
Gives correct answer: [‘Question: How many cats are there? Answer: Two\n’]
+However, when I change prompt to
+prompt2 = ""Question: How many dogs are there? Answer: ""
+inputs2 = processor(images=image, text=prompt2, return_tensors=“pt”).to(
+device, torch.float16
+)
outputs2 = model.generate(**inputs2)
+text2 = processor.tokenizer.batch_decode(outputs2, skip_special_tokens=True)
+print(text2)
[‘Question: How many dogs are there? Answer: Two\n’]
","Yeah. For example, CLIP can perfectly classify dogs and cats, but BLIP seems utterly unsuitable for classification…
" +Prevent creation of multiple checkpoints,https://discuss.huggingface.co/t/prevent-creation-of-multiple-checkpoints/168144,168144,5,2025-09-05 20:15:07.934000+00:00,"[{'id': 241309, 'name': 'Alex', 'username': 'SuperBowser', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9f8e36/{size}.png', 'created_at': '2025-09-05T20:15:08.005Z', 'cooked': 'In my training arguments I selected to save every 200 steps, but my model is fairly large (relative to my disk size). I would like to save every 200 steps, but every save should just overwrite previous save instead of creating new save point. Is this possible?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-09-05T20:15:08.005Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 5, 'readers_count': 4, 'score': 51.0, 'yours': False, 'topic_id': 168144, 'topic_slug': 'prevent-creation-of-multiple-checkpoints', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/prevent-creation-of-multiple-checkpoints/168144/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241317, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-06T00:19:59.432Z', 'cooked': 'Strictly speaking, it’s not overwriting, but I think save_total_limit or save_only_model are closer to the intended purpose.
from transformers import TrainingArguments\n\nargs = TrainingArguments(\n output_dir=""out"",\n save_strategy=""steps"",\n save_steps=200,\n save_total_limit=1, # deletes older checkpoints\n save_only_model=True, # 4.37+; skips optimizer/scheduler to shrink size\n)\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-09-08T01:48:01.261Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168144, 'topic_slug': 'prevent-creation-of-multiple-checkpoints', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/prevent-creation-of-multiple-checkpoints/168144/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","In my training arguments I selected to save every 200 steps, but my model is fairly large (relative to my disk size). I would like to save every 200 steps, but every save should just overwrite previous save instead of creating new save point. Is this possible?
","Strictly speaking, it’s not overwriting, but I think save_total_limit or save_only_model are closer to the intended purpose.
from transformers import TrainingArguments
+
+args = TrainingArguments(
+ output_dir=""out"",
+ save_strategy=""steps"",
+ save_steps=200,
+ save_total_limit=1, # deletes older checkpoints
+ save_only_model=True, # 4.37+; skips optimizer/scheduler to shrink size
+)
+Hi there,
\nI want to setup a LLM workstation to start developing my own agent and tools and experiment. I travel a lot and don’t have a big budget at the moment to expend.
\nI saw the Nvidia Jetson Nano Orin Super and it looks cool but I’m not sure if is the best option for my needs.
\nI use Linux and like to have freedom and don’t be tied to an specific ecosystem, there are very little reviews about this one and none of then cover Agentic development on deep.
I also read that a NVIDIA 3060 should be enough for my needs but I would have to use it as eGPU which has a shitty performance or build a mini workstation, which is a very attractive option and I wouldn’t mind to expend a bit more of money if it truly fits my needs.
\nSo what do I need/want??
\nI want to be able to develop agents and integrate them via CLI for Sysadmin and Cyber Security purposes, I would like to have a decent level of inference to basically play and explore as much is possible to know exactly what I will need in the future and develop tools that will scale once I have a more beefy setup.
\nI’m also interesting on coding agents but I guess I would need the capacity to train the model to achieve what i have in mind. And I don’t know how realistic it is to expect to be able to train model with such a low budget. At least I would like to run something that allows me to get ride of Cursor.
\nI really want to get my hands on ASAP but I’m afraid to make an investment that I will end regretting after I dive on LLMs more, that’s why I’m writing this post so maybe I can get some feedback and guidance about the best way to start this project based of my circumstances and needs
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-06T14:25:48.814Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 9, 'readers_count': 8, 'score': 41.8, 'yours': False, 'topic_id': 168164, 'topic_slug': 'low-budge-worstation', 'display_username': 'Nick Dandolos', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103255, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/low-budge-worstation/168164/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241381, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-06T22:54:16.714Z', 'cooked': 'For hardware consultations or fine-tuning, I think it’s best to ask questions on the HF Discord or Unsloth’s Discord.
\n\n\nNvidia Jetson Nano Orin Super and it looks cool but I’m not sure if is the best option for my needs.
\n
It’s cool but not well-suited for various tasks with LLM. It’s more geared toward edge devices, so I think it’s better to choose a GPU this time.
\n\n\na NVIDIA 3060 should be enough
\n
Yeah. I’m using a 3060 Ti too. Well, with 8GB of VRAM, you can manage some things. Ideally, 12GB or 16GB—the more VRAM you have, the more you can do. For anything other than high-end, VRAM size matters more than clock speed.
\n\n\nhow realistic it is to expect to be able to train model with such a low budget.
\n
I think this might be helpful.
\nBTW, setting aside security concerns, renting cloud GPUs for fine-tuning is straightforward. Google Colab, for instance.
\nWow, all this is awesome! Thank you very much!! I did also wrote this post on the Discord Server!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-06T23:16:12.784Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 168164, 'topic_slug': 'low-budge-worstation', 'display_username': 'Nick Dandolos', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103255, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/low-budge-worstation/168164/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241405, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-07T11:16:18.060Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-09-07T11:16:18.060Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 168164, 'topic_slug': 'low-budge-worstation', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/low-budge-worstation/168164/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi there,
+I want to setup a LLM workstation to start developing my own agent and tools and experiment. I travel a lot and don’t have a big budget at the moment to expend.
+I saw the Nvidia Jetson Nano Orin Super and it looks cool but I’m not sure if is the best option for my needs.
+I use Linux and like to have freedom and don’t be tied to an specific ecosystem, there are very little reviews about this one and none of then cover Agentic development on deep.
I also read that a NVIDIA 3060 should be enough for my needs but I would have to use it as eGPU which has a shitty performance or build a mini workstation, which is a very attractive option and I wouldn’t mind to expend a bit more of money if it truly fits my needs.
+So what do I need/want??
+I want to be able to develop agents and integrate them via CLI for Sysadmin and Cyber Security purposes, I would like to have a decent level of inference to basically play and explore as much is possible to know exactly what I will need in the future and develop tools that will scale once I have a more beefy setup.
+I’m also interesting on coding agents but I guess I would need the capacity to train the model to achieve what i have in mind. And I don’t know how realistic it is to expect to be able to train model with such a low budget. At least I would like to run something that allows me to get ride of Cursor.
+I really want to get my hands on ASAP but I’m afraid to make an investment that I will end regretting after I dive on LLMs more, that’s why I’m writing this post so maybe I can get some feedback and guidance about the best way to start this project based of my circumstances and needs
","For hardware consultations or fine-tuning, I think it’s best to ask questions on the HF Discord or Unsloth’s Discord.
+++Nvidia Jetson Nano Orin Super and it looks cool but I’m not sure if is the best option for my needs.
+
It’s cool but not well-suited for various tasks with LLM. It’s more geared toward edge devices, so I think it’s better to choose a GPU this time.
+++a NVIDIA 3060 should be enough
+
Yeah. I’m using a 3060 Ti too. Well, with 8GB of VRAM, you can manage some things. Ideally, 12GB or 16GB—the more VRAM you have, the more you can do. For anything other than high-end, VRAM size matters more than clock speed.
+++how realistic it is to expect to be able to train model with such a low budget.
+
I think this might be helpful.
+BTW, setting aside security concerns, renting cloud GPUs for fine-tuning is straightforward. Google Colab, for instance.
+Hi all,
\nI am trying to train a custom model for NLP sequence classification (multiclass) and struggling to be able to train it for a reason I don’t know, that is the reason why I am asking on this forum. I already had a look at similar posts on the forum with no luck.
\nFirst of all, my dataset looks like the following in DataFrame before introducing it to a dataset (5 instances per class or label, being 0 the lowest label number and 251 the maximum one, so 252 labels in total):
\n text label\n0 Configuración del área de selección de TV Set 0\n1 Configuración del área de selección de TV Set 0\n2 Conformación de la sección de selección de TV... 0\n3 Conformación ae la stcción de seldcción de TV Set 0\n4 Validar la configuración del área de selección... 0\n... ... ...\n1281 Validación incorrecta por identificador de art... 251\n1282 Validación incorrecta mediante identificador d... 251\n1283 Validación incorrecta por identificador de art... 251\n1284 Validación incorrecta por identificador de art... 251\n1285 Validar Validación incorrecta por identificado... 251\nAs It is a custom model, I changed the value of out_features at out_proj in the classification part, so the resulting architecture looks like the following:
\nRobertaForSequenceClassification(\n (roberta): RobertaModel(\n (embeddings): RobertaEmbeddings(\n (word_embeddings): Embedding(50262, 1024, padding_idx=1)\n (position_embeddings): Embedding(514, 1024, padding_idx=1)\n (token_type_embeddings): Embedding(1, 1024)\n (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n (dropout): Dropout(p=0.0, inplace=False)\n )\n (encoder): RobertaEncoder(\n (layer): ModuleList(\n (0-23): 24 x RobertaLayer(\n (attention): RobertaAttention(\n (self): RobertaSdpaSelfAttention(\n (query): Linear(in_features=1024, out_features=1024, bias=True)\n (key): Linear(in_features=1024, out_features=1024, bias=True)\n (value): Linear(in_features=1024, out_features=1024, bias=True)\n (dropout): Dropout(p=0.0, inplace=False)\n )\n (output): RobertaSelfOutput(\n (dense): Linear(in_features=1024, out_features=1024, bias=True)\n (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n (dropout): Dropout(p=0.0, inplace=False)\n )\n )\n (intermediate): RobertaIntermediate(\n (dense): Linear(in_features=1024, out_features=4096, bias=True)\n (intermediate_act_fn): GELUActivation()\n )\n (output): RobertaOutput(\n (dense): Linear(in_features=4096, out_features=1024, bias=True)\n (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n (dropout): Dropout(p=0.0, inplace=False)\n )\n )\n )\n )\n )\n (classifier): RobertaClassificationHead(\n (dense): Linear(in_features=1024, out_features=1024, bias=True)\n (dropout): Dropout(p=0.0, inplace=False)\n (out_proj): Linear(in_features=1024, out_features=252, bias=True)\n )\n)\nThen I use the following code in order to create a HuggingFace Dataset:
\ndataset = Dataset.from_pandas(df, split=\'train\')\ndataset = dataset.train_test_split(shuffle=True, seed=42, test_size=0.2)\nprint(dataset)\nWhere the print gives the following result (I already checked that values in label go from 0 to N-1 labels or classes):
\nDatasetDict({\n train: Dataset({\n features: [\'text\', \'label\'],\n num_rows: 1028\n })\n test: Dataset({\n features: [\'text\', \'label\'],\n num_rows: 258\n })\n})\nDespite having done all the remaining steps before training correctly (or so I believe) and having at least one instance per class in train and test dataset, when I get to the function train, I get the following error:
\n---------------------------------------------------------------------------\nIndexError Traceback (most recent call last)\nCell In[103], line 1\n----> 1 trainer.train()\n 2 modelo_peft.to(\'cpu\')\n 3 modelo_peft.eval()\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\transformers\\trainer.py:2238, in Trainer.train(self, resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs)\n 2236 hf_hub_utils.enable_progress_bars()\n 2237 else:\n-> 2238 return inner_training_loop(\n 2239 args=args,\n 2240 resume_from_checkpoint=resume_from_checkpoint,\n 2241 trial=trial,\n 2242 ignore_keys_for_eval=ignore_keys_for_eval,\n 2243 )\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\transformers\\trainer.py:2582, in Trainer._inner_training_loop(self, batch_size, args, resume_from_checkpoint, trial, ignore_keys_for_eval)\n 2575 context = (\n 2576 functools.partial(self.accelerator.no_sync, model=model)\n 2577 if i != len(batch_samples) - 1\n 2578 and self.accelerator.distributed_type != DistributedType.DEEPSPEED\n 2579 else contextlib.nullcontext\n 2580 )\n 2581 with context():\n-> 2582 tr_loss_step = self.training_step(model, inputs, num_items_in_batch)\n 2584 if (\n 2585 args.logging_nan_inf_filter\n 2586 and not is_torch_xla_available()\n 2587 and (torch.isnan(tr_loss_step) or torch.isinf(tr_loss_step))\n 2588 ):\n 2589 # if loss is nan or inf simply add the average of previous logged losses\n 2590 tr_loss = tr_loss + tr_loss / (1 + self.state.global_step - self._globalstep_last_logged)\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\transformers\\trainer.py:3796, in Trainer.training_step(self, model, inputs, num_items_in_batch)\n 3793 return loss_mb.reduce_mean().detach().to(self.args.device)\n 3795 with self.compute_loss_context_manager():\n-> 3796 loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch)\n 3798 del inputs\n 3799 if (\n 3800 self.args.torch_empty_cache_steps is not None\n 3801 and self.state.global_step % self.args.torch_empty_cache_steps == 0\n 3802 ):\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\transformers\\trainer.py:3884, in Trainer.compute_loss(self, model, inputs, return_outputs, num_items_in_batch)\n 3882 kwargs[""num_items_in_batch""] = num_items_in_batch\n 3883 inputs = {**inputs, **kwargs}\n-> 3884 outputs = model(**inputs)\n 3885 # Save past state if it exists\n 3886 # TODO: this needs to be fixed and made cleaner later.\n 3887 if self.args.past_index >= 0:\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773, in Module._wrapped_call_impl(self, *args, **kwargs)\n 1771 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]\n 1772 else:\n-> 1773 return self._call_impl(*args, **kwargs)\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784, in Module._call_impl(self, *args, **kwargs)\n 1779 # If we don\'t have any hooks, we want to skip the rest of the logic in\n 1780 # this function, and just call forward.\n 1781 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks\n 1782 or _global_backward_pre_hooks or _global_backward_hooks\n 1783 or _global_forward_hooks or _global_forward_pre_hooks):\n-> 1784 return forward_call(*args, **kwargs)\n 1786 result = None\n 1787 called_always_called_hooks = set()\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\peft\\peft_model.py:1652, in PeftModelForSequenceClassification.forward(self, input_ids, attention_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict, task_ids, **kwargs)\n 1650 if peft_config.peft_type == PeftType.POLY:\n 1651 kwargs[""task_ids""] = task_ids\n-> 1652 return self.base_model(\n 1653 input_ids=input_ids,\n 1654 attention_mask=attention_mask,\n 1655 inputs_embeds=inputs_embeds,\n 1656 labels=labels,\n 1657 output_attentions=output_attentions,\n 1658 output_hidden_states=output_hidden_states,\n 1659 return_dict=return_dict,\n 1660 **kwargs,\n 1661 )\n 1663 batch_size = _get_batch_size(input_ids, inputs_embeds)\n 1664 if attention_mask is not None:\n 1665 # concat prompt attention mask\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773, in Module._wrapped_call_impl(self, *args, **kwargs)\n 1771 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]\n 1772 else:\n-> 1773 return self._call_impl(*args, **kwargs)\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784, in Module._call_impl(self, *args, **kwargs)\n 1779 # If we don\'t have any hooks, we want to skip the rest of the logic in\n 1780 # this function, and just call forward.\n 1781 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks\n 1782 or _global_backward_pre_hooks or _global_backward_hooks\n 1783 or _global_forward_hooks or _global_forward_pre_hooks):\n-> 1784 return forward_call(*args, **kwargs)\n 1786 result = None\n 1787 called_always_called_hooks = set()\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\peft\\tuners\\tuners_utils.py:222, in BaseTuner.forward(self, *args, **kwargs)\n 221 def forward(self, *args: Any, **kwargs: Any):\n--> 222 return self.model.forward(*args, **kwargs)\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\transformers\\models\\roberta\\modeling_roberta.py:1228, in RobertaForSequenceClassification.forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict)\n 1226 elif self.config.problem_type == ""single_label_classification"":\n 1227 loss_fct = CrossEntropyLoss()\n-> 1228 loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))\n 1229 elif self.config.problem_type == ""multi_label_classification"":\n 1230 loss_fct = BCEWithLogitsLoss()\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773, in Module._wrapped_call_impl(self, *args, **kwargs)\n 1771 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]\n 1772 else:\n-> 1773 return self._call_impl(*args, **kwargs)\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784, in Module._call_impl(self, *args, **kwargs)\n 1779 # If we don\'t have any hooks, we want to skip the rest of the logic in\n 1780 # this function, and just call forward.\n 1781 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks\n 1782 or _global_backward_pre_hooks or _global_backward_hooks\n 1783 or _global_forward_hooks or _global_forward_pre_hooks):\n-> 1784 return forward_call(*args, **kwargs)\n 1786 result = None\n 1787 called_always_called_hooks = set()\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\modules\\loss.py:1310, in CrossEntropyLoss.forward(self, input, target)\n 1309 def forward(self, input: Tensor, target: Tensor) -> Tensor:\n-> 1310 return F.cross_entropy(\n 1311 input,\n 1312 target,\n 1313 weight=self.weight,\n 1314 ignore_index=self.ignore_index,\n 1315 reduction=self.reduction,\n 1316 label_smoothing=self.label_smoothing,\n 1317 )\n\nFile ~\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\torch\\nn\\functional.py:3462, in cross_entropy(input, target, weight, size_average, ignore_index, reduce, reduction, label_smoothing)\n 3460 if size_average is not None or reduce is not None:\n 3461 reduction = _Reduction.legacy_get_string(size_average, reduce)\n-> 3462 return torch._C._nn.cross_entropy_loss(\n 3463 input,\n 3464 target,\n 3465 weight,\n 3466 _Reduction.get_enum(reduction),\n 3467 ignore_index,\n 3468 label_smoothing,\n 3469 )\n\nIndexError: Target 134 is out of bounds.\nAny ideas of what may be wrong? Let me know if any other information is needed.
\nThanks,
\nJavier
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-06T10:35:54.160Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 7, 'readers_count': 6, 'score': 41.4, 'yours': False, 'topic_id': 168143, 'topic_slug': 'indexerror-target-n-is-out-of-bounds-within-trainer-train-function', 'display_username': 'Javier M.A.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 4, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103219, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/indexerror-target-n-is-out-of-bounds-within-trainer-train-function/168143/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241316, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-06T00:10:31.575Z', 'cooked': 'This may occur if num_labels is not passed during model loading.
from datasets import Dataset\nfrom transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments\nimport numpy as np\nimport pandas as pd\nimport torch\nimport math\n\n# 0) Example dataframe (replace with your df)\n# df = pd.read_csv(""your_data.csv"") # must contain \'text\' and integer \'label\'\ndf = pd.DataFrame({\n ""text"": [f""ejemplo {i}"" for i in range(3000)],\n ""label"": np.repeat(np.arange(252), repeats=math.ceil(3000/252))[:3000]\n})\n\n# 1) Ensure labels are 0..C-1\nC = int(df[""label""].max() + 1)\nm = int(df[""label""].min())\nif m != 0:\n df[""label""] = df[""label""] - m\nassert df[""label""].between(0, C - 1).all(), ""labels must be in [0, C-1]""\n\n# 2) Build small train/test datasets\nds = Dataset.from_pandas(df[[""text"", ""label""]], split=""train"").train_test_split(test_size=0.1, seed=42)\n\n# 3) Tokenize\ntok = AutoTokenizer.from_pretrained(""roberta-base"")\ndef preprocess(ex):\n return tok(ex[""text""], truncation=True, padding=""max_length"", max_length=64)\nds_tok = ds.map(preprocess, batched=True).remove_columns([""text""]).with_format(""torch"")\n\n# 4) Create model with the correct class count; let Transformers swap the head\nmodel = AutoModelForSequenceClassification.from_pretrained(\n ""roberta-base"",\n num_labels=C, # tells the new classifier size\n ignore_mismatched_sizes=True, # skip loading the old head\n)\n# optional but recommended: explicit label maps\nmodel.config.id2label = {i: str(i) for i in range(C)}\nmodel.config.label2id = {v: k for k, v in model.config.id2label.items()}\n\n# 5) Train briefly\nargs = TrainingArguments(\n output_dir=""out_fix"",\n per_device_train_batch_size=8,\n per_device_eval_batch_size=8,\n learning_rate=5e-5,\n num_train_epochs=1,\n logging_steps=10,\n eval_strategy=""no"",\n report_to=""none"",\n)\n\ntrainer = Trainer(model=model, args=args, train_dataset=ds_tok[""train""])\ntrainer.train() # IndexError: Target ** is out of bounds. (If without num_labels and ignore_mismatched_sizes)\nMany thanks for your answer John. Regarding what you said regarding num_labels, the way I did it in my code was the following (first line in the following code):
\nnueva_configuracion_modelo = AutoConfig.from_pretrained(nombre_modelo, num_labels=numero_de_etiquetas, id2label=ids_a_etiquetas, label2id=etiquetas_a_id, cache_dir=\'./huggingface_mirror\')\n\nmodelo_roberta = AutoModelForSequenceClassification.from_pretrained(\'PlanTL-GOB-ES/roberta-large-bne-massive\', cache_dir=\'./huggingface_mirror\', local_files_only=True)\n\n\nif modelo_roberta.config.num_labels != nueva_configuracion_modelo.num_labels or modelo_roberta.config.id2label != nueva_configuracion_modelo_config.id2label:\n modelo_roberta.classifier.out_proj.out_features=nueva_configuracion_modelo.num_labels\n \nmodelo_roberta.config = nueva_configuracion_modelo\n\nprint(modelo_roberta.config)\n\ntokenizador_roberta = AutoTokenizer.from_pretrained(nombre_modelo, cache_dir=\'./huggingface_mirror\', local_files_only=True, from_pt=True)\nWith that code I changed the value in out_features parameter of layer out_proj in the classification part to 252 (the number of different classes) and saw label2id and id2label updated with values from my custom model.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-06T11:12:36.335Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 168143, 'topic_slug': 'indexerror-target-n-is-out-of-bounds-within-trainer-train-function', 'display_username': 'Javier M.A.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103219, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/indexerror-target-n-is-out-of-bounds-within-trainer-train-function/168143/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241348, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-06T13:12:56.958Z', 'cooked': 'In that case, the actual weigh probably won’t change t even if the attribute is modified.
\nfrom transformers import AutoModelForSequenceClassification, AutoTokenizer\nimport torch\n\n# 1) Load a small model with 2 labels so the classifier head is tiny\nmodel = AutoModelForSequenceClassification.from_pretrained(""roberta-base"", num_labels=2)\ntok = AutoTokenizer.from_pretrained(""roberta-base"")\n\nhead = model.classifier.out_proj # this is an nn.Linear\n\nprint(""=== BEFORE ==="")\nprint(""repr:"", head)\nprint(""out_features attr:"", head.out_features)\nprint(""weight shape:"", tuple(head.weight.shape))\nprint(""bias shape:"", tuple(head.bias.shape))\n\n# 2) Change ONLY the attribute (what your code effectively does)\nhead.out_features = 252 # <-- attribute changed, tensors untouched\n\nprint(""\\n=== AFTER CHANGING ATTRIBUTE ONLY ==="")\nprint(""repr:"", head) # repr now claims out_features=252\nprint(""out_features attr:"", head.out_features)\nprint(""weight shape:"", tuple(head.weight.shape)) # still (2, hidden_size)\nprint(""bias shape:"", tuple(head.bias.shape)) # still (2,)\n\n# 3) Show the model still produces 2 logits, not 252\nbatch = tok(""hola mundo"", return_tensors=""pt"", padding=True, truncation=True, max_length=16)\nwith torch.no_grad():\n logits = model(**batch).logits\nprint(""\\nlogits shape from forward():"", tuple(logits.shape)) # last dim is 2\n\n# 4) The correct fix is to REPLACE the Linear layer\nin_f = head.in_features\nmodel.classifier.out_proj = torch.nn.Linear(in_f, 252, bias=True)\n\nprint(""\\n=== AFTER REPLACING THE LAYER ==="")\nprint(""repr:"", model.classifier.out_proj)\nprint(""out_features attr:"", model.classifier.out_proj.out_features)\nprint(""weight shape:"", tuple(model.classifier.out_proj.weight.shape)) # now (252, hidden_size)\nprint(""bias shape:"", tuple(model.classifier.out_proj.bias.shape)) # now (252,)\n\nwith torch.no_grad():\n logits = model(**batch).logits\nprint(""logits shape from forward():"", tuple(logits.shape)) # last dim is 252\n""""""\n=== BEFORE ===\nrepr: Linear(in_features=768, out_features=2, bias=True)\nout_features attr: 2\nweight shape: (2, 768)\nbias shape: (2,)\n\n=== AFTER CHANGING ATTRIBUTE ONLY ===\nrepr: Linear(in_features=768, out_features=252, bias=True)\nout_features attr: 252\nweight shape: (2, 768)\nbias shape: (2,)\n\nlogits shape from forward(): (1, 2)\n\n=== AFTER REPLACING THE LAYER ===\nrepr: Linear(in_features=768, out_features=252, bias=True)\nout_features attr: 252\nweight shape: (252, 768)\nbias shape: (252,)\nlogits shape from forward(): (1, 252)\n""""""\nYou were totally right John ! I just printed the weight and bias in my code and the results were the original ones, so indeed I was modifying it the wrong way.
So following the example I modified my code from this:
\nif modelo_roberta.config.num_labels != nueva_configuracion_modelo.num_labels or modelo_roberta.config.id2label != nueva_configuracion_modelo_config.id2label:\n modelo_roberta.classifier.out_proj.out_features=nueva_configuracion_modelo.num_labels\n \nmodelo_roberta.config = nueva_configuracion_modelo\nTo this:
\nmodelo_roberta.classifier.out_proj = torch.nn.Linear(modelo_roberta.classifier.out_proj.in_features, numero_de_etiquetas, bias=True)\nmodelo_roberta.num_labels = numero_de_etiquetas\nmodelo_roberta.config = nueva_configuracion_modelo\nAnd now it trains.
\nMany thanks for your help!
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-09-06T16:35:51.006Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 168143, 'topic_slug': 'indexerror-target-n-is-out-of-bounds-within-trainer-train-function', 'display_username': 'Javier M.A.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103219, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/indexerror-target-n-is-out-of-bounds-within-trainer-train-function/168143/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241392, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-07T04:13:52.319Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-09-07T04:13:52.319Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168143, 'topic_slug': 'indexerror-target-n-is-out-of-bounds-within-trainer-train-function', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/indexerror-target-n-is-out-of-bounds-within-trainer-train-function/168143/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi all,
+I am trying to train a custom model for NLP sequence classification (multiclass) and struggling to be able to train it for a reason I don’t know, that is the reason why I am asking on this forum. I already had a look at similar posts on the forum with no luck.
+First of all, my dataset looks like the following in DataFrame before introducing it to a dataset (5 instances per class or label, being 0 the lowest label number and 251 the maximum one, so 252 labels in total):
+ text label
+0 Configuración del área de selección de TV Set 0
+1 Configuración del área de selección de TV Set 0
+2 Conformación de la sección de selección de TV... 0
+3 Conformación ae la stcción de seldcción de TV Set 0
+4 Validar la configuración del área de selección... 0
+... ... ...
+1281 Validación incorrecta por identificador de art... 251
+1282 Validación incorrecta mediante identificador d... 251
+1283 Validación incorrecta por identificador de art... 251
+1284 Validación incorrecta por identificador de art... 251
+1285 Validar Validación incorrecta por identificado... 251
+As It is a custom model, I changed the value of out_features at out_proj in the classification part, so the resulting architecture looks like the following:
+RobertaForSequenceClassification(
+ (roberta): RobertaModel(
+ (embeddings): RobertaEmbeddings(
+ (word_embeddings): Embedding(50262, 1024, padding_idx=1)
+ (position_embeddings): Embedding(514, 1024, padding_idx=1)
+ (token_type_embeddings): Embedding(1, 1024)
+ (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
+ (dropout): Dropout(p=0.0, inplace=False)
+ )
+ (encoder): RobertaEncoder(
+ (layer): ModuleList(
+ (0-23): 24 x RobertaLayer(
+ (attention): RobertaAttention(
+ (self): RobertaSdpaSelfAttention(
+ (query): Linear(in_features=1024, out_features=1024, bias=True)
+ (key): Linear(in_features=1024, out_features=1024, bias=True)
+ (value): Linear(in_features=1024, out_features=1024, bias=True)
+ (dropout): Dropout(p=0.0, inplace=False)
+ )
+ (output): RobertaSelfOutput(
+ (dense): Linear(in_features=1024, out_features=1024, bias=True)
+ (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
+ (dropout): Dropout(p=0.0, inplace=False)
+ )
+ )
+ (intermediate): RobertaIntermediate(
+ (dense): Linear(in_features=1024, out_features=4096, bias=True)
+ (intermediate_act_fn): GELUActivation()
+ )
+ (output): RobertaOutput(
+ (dense): Linear(in_features=4096, out_features=1024, bias=True)
+ (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
+ (dropout): Dropout(p=0.0, inplace=False)
+ )
+ )
+ )
+ )
+ )
+ (classifier): RobertaClassificationHead(
+ (dense): Linear(in_features=1024, out_features=1024, bias=True)
+ (dropout): Dropout(p=0.0, inplace=False)
+ (out_proj): Linear(in_features=1024, out_features=252, bias=True)
+ )
+)
+Then I use the following code in order to create a HuggingFace Dataset:
+dataset = Dataset.from_pandas(df, split='train')
+dataset = dataset.train_test_split(shuffle=True, seed=42, test_size=0.2)
+print(dataset)
+Where the print gives the following result (I already checked that values in label go from 0 to N-1 labels or classes):
+DatasetDict({
+ train: Dataset({
+ features: ['text', 'label'],
+ num_rows: 1028
+ })
+ test: Dataset({
+ features: ['text', 'label'],
+ num_rows: 258
+ })
+})
+Despite having done all the remaining steps before training correctly (or so I believe) and having at least one instance per class in train and test dataset, when I get to the function train, I get the following error:
+---------------------------------------------------------------------------
+IndexError Traceback (most recent call last)
+Cell In[103], line 1
+----> 1 trainer.train()
+ 2 modelo_peft.to('cpu')
+ 3 modelo_peft.eval()
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\transformers\trainer.py:2238, in Trainer.train(self, resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs)
+ 2236 hf_hub_utils.enable_progress_bars()
+ 2237 else:
+-> 2238 return inner_training_loop(
+ 2239 args=args,
+ 2240 resume_from_checkpoint=resume_from_checkpoint,
+ 2241 trial=trial,
+ 2242 ignore_keys_for_eval=ignore_keys_for_eval,
+ 2243 )
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\transformers\trainer.py:2582, in Trainer._inner_training_loop(self, batch_size, args, resume_from_checkpoint, trial, ignore_keys_for_eval)
+ 2575 context = (
+ 2576 functools.partial(self.accelerator.no_sync, model=model)
+ 2577 if i != len(batch_samples) - 1
+ 2578 and self.accelerator.distributed_type != DistributedType.DEEPSPEED
+ 2579 else contextlib.nullcontext
+ 2580 )
+ 2581 with context():
+-> 2582 tr_loss_step = self.training_step(model, inputs, num_items_in_batch)
+ 2584 if (
+ 2585 args.logging_nan_inf_filter
+ 2586 and not is_torch_xla_available()
+ 2587 and (torch.isnan(tr_loss_step) or torch.isinf(tr_loss_step))
+ 2588 ):
+ 2589 # if loss is nan or inf simply add the average of previous logged losses
+ 2590 tr_loss = tr_loss + tr_loss / (1 + self.state.global_step - self._globalstep_last_logged)
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\transformers\trainer.py:3796, in Trainer.training_step(self, model, inputs, num_items_in_batch)
+ 3793 return loss_mb.reduce_mean().detach().to(self.args.device)
+ 3795 with self.compute_loss_context_manager():
+-> 3796 loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch)
+ 3798 del inputs
+ 3799 if (
+ 3800 self.args.torch_empty_cache_steps is not None
+ 3801 and self.state.global_step % self.args.torch_empty_cache_steps == 0
+ 3802 ):
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\transformers\trainer.py:3884, in Trainer.compute_loss(self, model, inputs, return_outputs, num_items_in_batch)
+ 3882 kwargs[""num_items_in_batch""] = num_items_in_batch
+ 3883 inputs = {**inputs, **kwargs}
+-> 3884 outputs = model(**inputs)
+ 3885 # Save past state if it exists
+ 3886 # TODO: this needs to be fixed and made cleaner later.
+ 3887 if self.args.past_index >= 0:
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\module.py:1773, in Module._wrapped_call_impl(self, *args, **kwargs)
+ 1771 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
+ 1772 else:
+-> 1773 return self._call_impl(*args, **kwargs)
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\module.py:1784, in Module._call_impl(self, *args, **kwargs)
+ 1779 # If we don't have any hooks, we want to skip the rest of the logic in
+ 1780 # this function, and just call forward.
+ 1781 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
+ 1782 or _global_backward_pre_hooks or _global_backward_hooks
+ 1783 or _global_forward_hooks or _global_forward_pre_hooks):
+-> 1784 return forward_call(*args, **kwargs)
+ 1786 result = None
+ 1787 called_always_called_hooks = set()
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\peft\peft_model.py:1652, in PeftModelForSequenceClassification.forward(self, input_ids, attention_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict, task_ids, **kwargs)
+ 1650 if peft_config.peft_type == PeftType.POLY:
+ 1651 kwargs[""task_ids""] = task_ids
+-> 1652 return self.base_model(
+ 1653 input_ids=input_ids,
+ 1654 attention_mask=attention_mask,
+ 1655 inputs_embeds=inputs_embeds,
+ 1656 labels=labels,
+ 1657 output_attentions=output_attentions,
+ 1658 output_hidden_states=output_hidden_states,
+ 1659 return_dict=return_dict,
+ 1660 **kwargs,
+ 1661 )
+ 1663 batch_size = _get_batch_size(input_ids, inputs_embeds)
+ 1664 if attention_mask is not None:
+ 1665 # concat prompt attention mask
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\module.py:1773, in Module._wrapped_call_impl(self, *args, **kwargs)
+ 1771 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
+ 1772 else:
+-> 1773 return self._call_impl(*args, **kwargs)
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\module.py:1784, in Module._call_impl(self, *args, **kwargs)
+ 1779 # If we don't have any hooks, we want to skip the rest of the logic in
+ 1780 # this function, and just call forward.
+ 1781 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
+ 1782 or _global_backward_pre_hooks or _global_backward_hooks
+ 1783 or _global_forward_hooks or _global_forward_pre_hooks):
+-> 1784 return forward_call(*args, **kwargs)
+ 1786 result = None
+ 1787 called_always_called_hooks = set()
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\peft\tuners\tuners_utils.py:222, in BaseTuner.forward(self, *args, **kwargs)
+ 221 def forward(self, *args: Any, **kwargs: Any):
+--> 222 return self.model.forward(*args, **kwargs)
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\transformers\models\roberta\modeling_roberta.py:1228, in RobertaForSequenceClassification.forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict)
+ 1226 elif self.config.problem_type == ""single_label_classification"":
+ 1227 loss_fct = CrossEntropyLoss()
+-> 1228 loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+ 1229 elif self.config.problem_type == ""multi_label_classification"":
+ 1230 loss_fct = BCEWithLogitsLoss()
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\module.py:1773, in Module._wrapped_call_impl(self, *args, **kwargs)
+ 1771 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
+ 1772 else:
+-> 1773 return self._call_impl(*args, **kwargs)
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\module.py:1784, in Module._call_impl(self, *args, **kwargs)
+ 1779 # If we don't have any hooks, we want to skip the rest of the logic in
+ 1780 # this function, and just call forward.
+ 1781 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
+ 1782 or _global_backward_pre_hooks or _global_backward_hooks
+ 1783 or _global_forward_hooks or _global_forward_pre_hooks):
+-> 1784 return forward_call(*args, **kwargs)
+ 1786 result = None
+ 1787 called_always_called_hooks = set()
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\loss.py:1310, in CrossEntropyLoss.forward(self, input, target)
+ 1309 def forward(self, input: Tensor, target: Tensor) -> Tensor:
+-> 1310 return F.cross_entropy(
+ 1311 input,
+ 1312 target,
+ 1313 weight=self.weight,
+ 1314 ignore_index=self.ignore_index,
+ 1315 reduction=self.reduction,
+ 1316 label_smoothing=self.label_smoothing,
+ 1317 )
+
+File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\functional.py:3462, in cross_entropy(input, target, weight, size_average, ignore_index, reduce, reduction, label_smoothing)
+ 3460 if size_average is not None or reduce is not None:
+ 3461 reduction = _Reduction.legacy_get_string(size_average, reduce)
+-> 3462 return torch._C._nn.cross_entropy_loss(
+ 3463 input,
+ 3464 target,
+ 3465 weight,
+ 3466 _Reduction.get_enum(reduction),
+ 3467 ignore_index,
+ 3468 label_smoothing,
+ 3469 )
+
+IndexError: Target 134 is out of bounds.
+Any ideas of what may be wrong? Let me know if any other information is needed.
+Thanks,
+Javier
","In that case, the actual weigh probably won’t change t even if the attribute is modified.
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
+import torch
+
+# 1) Load a small model with 2 labels so the classifier head is tiny
+model = AutoModelForSequenceClassification.from_pretrained(""roberta-base"", num_labels=2)
+tok = AutoTokenizer.from_pretrained(""roberta-base"")
+
+head = model.classifier.out_proj # this is an nn.Linear
+
+print(""=== BEFORE ==="")
+print(""repr:"", head)
+print(""out_features attr:"", head.out_features)
+print(""weight shape:"", tuple(head.weight.shape))
+print(""bias shape:"", tuple(head.bias.shape))
+
+# 2) Change ONLY the attribute (what your code effectively does)
+head.out_features = 252 # <-- attribute changed, tensors untouched
+
+print(""\n=== AFTER CHANGING ATTRIBUTE ONLY ==="")
+print(""repr:"", head) # repr now claims out_features=252
+print(""out_features attr:"", head.out_features)
+print(""weight shape:"", tuple(head.weight.shape)) # still (2, hidden_size)
+print(""bias shape:"", tuple(head.bias.shape)) # still (2,)
+
+# 3) Show the model still produces 2 logits, not 252
+batch = tok(""hola mundo"", return_tensors=""pt"", padding=True, truncation=True, max_length=16)
+with torch.no_grad():
+ logits = model(**batch).logits
+print(""\nlogits shape from forward():"", tuple(logits.shape)) # last dim is 2
+
+# 4) The correct fix is to REPLACE the Linear layer
+in_f = head.in_features
+model.classifier.out_proj = torch.nn.Linear(in_f, 252, bias=True)
+
+print(""\n=== AFTER REPLACING THE LAYER ==="")
+print(""repr:"", model.classifier.out_proj)
+print(""out_features attr:"", model.classifier.out_proj.out_features)
+print(""weight shape:"", tuple(model.classifier.out_proj.weight.shape)) # now (252, hidden_size)
+print(""bias shape:"", tuple(model.classifier.out_proj.bias.shape)) # now (252,)
+
+with torch.no_grad():
+ logits = model(**batch).logits
+print(""logits shape from forward():"", tuple(logits.shape)) # last dim is 252
+""""""
+=== BEFORE ===
+repr: Linear(in_features=768, out_features=2, bias=True)
+out_features attr: 2
+weight shape: (2, 768)
+bias shape: (2,)
+
+=== AFTER CHANGING ATTRIBUTE ONLY ===
+repr: Linear(in_features=768, out_features=252, bias=True)
+out_features attr: 252
+weight shape: (2, 768)
+bias shape: (2,)
+
+logits shape from forward(): (1, 2)
+
+=== AFTER REPLACING THE LAYER ===
+repr: Linear(in_features=768, out_features=252, bias=True)
+out_features attr: 252
+weight shape: (252, 768)
+bias shape: (252,)
+logits shape from forward(): (1, 252)
+""""""
+The following code produces error:
\nfrom transformers import AutoModelForSequenceClassification\nmodel_name = ‘openai/gpt-oss-20b’\nmodel = AutoModelForSequenceClassification.from_pretrained(model_name)\nError:
\nValueError:\n Unrecognized configuration class <class \'transformers.models.gpt_oss.configuration_gpt_oss.GptOssConfig\'> for this kind of \nAutoModel: AutoModelForSequenceClassification.\nMy transformers._version_ = 4.55.4
\nHere is full trace:
\n\n--------------------------------------------------------------------------- \nValueError Traceback (most recent call last) \n/tmp/ipython-input-2075936628.py in <cell line: 0>() 1 from transformers import AutoModelForSequenceClassification\n 2 model_name = \'openai/gpt-oss-20b\' \n----> 3 model = AutoModelForSequenceClassification.from_pretrained(model_name) \n/usr/local/lib/python3.12/dist-packages/transformers/models/auto/auto_factory.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs) 601 pretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, **kwargs 602 ) \n--> 603 raise ValueError( \n604 f""Unrecognized configuration class {config.__class__} for this kind of AutoModel: {cls.__name__}.\\n"" \n605 f""Model type should be one of {\', \'.join(c.__name__ for c in cls._model_mapping)}."" \nValueError: Unrecognized configuration class <class \'transformers.models.gpt_oss.configuration_gpt_oss.GptOssConfig\'> for this kind of AutoModel: AutoModelForSequenceClassification. Model type should be one of \nAlbertConfig, ArceeConfig, BartConfig, BertConfig, BigBirdConfig, BigBirdPegasusConfig, BioGptConfig, BloomConfig, CamembertConfig, CanineConfig, \nLlamaConfig, ConvBertConfig, CTRLConfig, Data2VecTextConfig, DebertaConfig, \nDebertaV2Config, DeepseekV2Config, DiffLlamaConfig, DistilBertConfig, \nDogeConfig, ElectraConfig, ErnieConfig, ErnieMConfig, EsmConfig, Exaone4Config, FalconConfig, FlaubertConfig, FNetConfig, FunnelConfig, GemmaConfig, Gemma2Config, Gemma3Config, GlmConfig, Glm4Config, GPT2Config, GPT2Config, \nGPTBigCodeConfig, GPTNeoConfig, GPTNeoXConfig, GPTJConfig, HeliumConfig, \nIBertConfig, JambaConfig, JetMoeConfig, LayoutLMConfig, LayoutLMv2Config, LayoutLMv3Config, LEDConfig, LiltConfig, LlamaConfig, LongformerConfig, \nLukeConfig, MarkupLMConfig, MBartConfig, MegaConfig, MegatronBertConfig, \nMiniMaxConfig, MistralConfig, MixtralConfig, MobileBertConfig, \nModernBertConfig, ModernBertDecoderConfig, MPNetConfig, MptConfig, MraConfig, \nMT5Config, MvpConfig, NemotronConfig, NezhaConfig, NystromformerConfig, OpenLlamaConfig, OpenAIGPTConfig, OPTConfig, PerceiverConfig, PersimmonConfig, PhiConfig, Phi3Config, PhimoeConfig, PLBartConfig, QDQBertConfig, Qwen2Config, \nQwen2MoeConfig, Qwen3Config, Qwen3MoeConfig, ReformerConfig, RemBertConfig, \nRobertaConfig, RobertaPreLayerNormConfig, RoCBertConfig, RoFormerConfig, \nSmolLM3Config, SqueezeBertConfig, StableLmConfig, Starcoder2Config, T5Config, T5GemmaConfig, TapasConfig, TransfoXLConfig, UMT5Config, XLMCon...\nIt seems to have just been implemented. GitHub version might work.
\npip install git+https://github.com/huggingface/transformers\nThank you so much again!
\nI need to download and later install this version of transformers offline.
\nHere is what I did:
\n!pip download git+https://github.com/huggingface/transformers -d ./wheels
and later I ran (offline) in Kaggle notebook:
\n!pip install wheels/transformers-4.57.0.dev0.zip
but it generated error:
\nProcessing ./wheels/transformers-4.57.0.dev0.zip\n error: subprocess-exited-with-error\n \n × pip subprocess to install build dependencies did not run successfully.\n │ exit code: 1\n ╰─> See above for output.\n \n note: This error originates from a subprocess, and is likely not a problem with pip.\n Installing build dependencies ... error\nerror: subprocess-exited-with-error\n\n× pip subprocess to install build dependencies did not run successfully.\n│ exit code: 1\n╰─> See above for output.\n\nnote: This error originates from a subprocess, and is likely not a problem with pip.\nIs it possible to download with dependencies and save?
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-09-03T20:04:43.284Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 1, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 167904, 'topic_slug': 'openai-gpt-oss-20b-what-heads-are-available', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/openai-gpt-oss-20b-what-heads-are-available/167904/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241136, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-09-03T23:21:34.755Z', 'cooked': 'For offline installation, you’ll probably need to use --no-index to avoid PyPI. Maybe like this?
# Online\n# Build a wheel from GitHub (avoid sdists)\ngit clone https://github.com/huggingface/transformers\ncd transformers\npython -m pip install -U build\npython -m build --wheel -o ../wheels\ncd ..\n# Offline\nWH=/kaggle/input/<your-dataset>/wheels\npip install --no-index --find-links=""$WH"" ""transformers==4.57.0.dev0""\nThank you so much!
\nWhen I run in Kaggle notebook !build --wheel -o ../wheels
I get back: /bin/bash: line 1: build: command not found
I also tried unsuccessfully
\n!python -m build --wheel -o ../wheels
Hmm, I might have forgotten to download build. I don’t know Kaggle…
# Online\n# Build a wheel from GitHub (avoid sdists)\ngit clone https://github.com/huggingface/transformers\ncd transformers\npython -m pip install -U build\npython -m build --wheel -o ../wheels\ncd ..\npython -m pip download --only-binary=:all: -d wheelhouse \\\n build setuptools wheel packaging pyproject_hooks setuptools-scm\n# Offline\nWH=/kaggle/input/<your-dataset>/wheels\npip install --no-index --find-links=""$WH"" \\\n build setuptools wheel packaging pyproject_hooks\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-09-05T12:50:18.113Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 167904, 'topic_slug': 'openai-gpt-oss-20b-what-heads-are-available', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/openai-gpt-oss-20b-what-heads-are-available/167904/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","The following code produces error:
+from transformers import AutoModelForSequenceClassification
+model_name = ‘openai/gpt-oss-20b’
+model = AutoModelForSequenceClassification.from_pretrained(model_name)
+Error:
+ValueError:
+ Unrecognized configuration class <class 'transformers.models.gpt_oss.configuration_gpt_oss.GptOssConfig'> for this kind of
+AutoModel: AutoModelForSequenceClassification.
+My transformers._version_ = 4.55.4
+Here is full trace:
+
+---------------------------------------------------------------------------
+ValueError Traceback (most recent call last)
+/tmp/ipython-input-2075936628.py in <cell line: 0>() 1 from transformers import AutoModelForSequenceClassification
+ 2 model_name = 'openai/gpt-oss-20b'
+----> 3 model = AutoModelForSequenceClassification.from_pretrained(model_name)
+/usr/local/lib/python3.12/dist-packages/transformers/models/auto/auto_factory.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs) 601 pretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, **kwargs 602 )
+--> 603 raise ValueError(
+604 f""Unrecognized configuration class {config.__class__} for this kind of AutoModel: {cls.__name__}.\n""
+605 f""Model type should be one of {', '.join(c.__name__ for c in cls._model_mapping)}.""
+ValueError: Unrecognized configuration class <class 'transformers.models.gpt_oss.configuration_gpt_oss.GptOssConfig'> for this kind of AutoModel: AutoModelForSequenceClassification. Model type should be one of
+AlbertConfig, ArceeConfig, BartConfig, BertConfig, BigBirdConfig, BigBirdPegasusConfig, BioGptConfig, BloomConfig, CamembertConfig, CanineConfig,
+LlamaConfig, ConvBertConfig, CTRLConfig, Data2VecTextConfig, DebertaConfig,
+DebertaV2Config, DeepseekV2Config, DiffLlamaConfig, DistilBertConfig,
+DogeConfig, ElectraConfig, ErnieConfig, ErnieMConfig, EsmConfig, Exaone4Config, FalconConfig, FlaubertConfig, FNetConfig, FunnelConfig, GemmaConfig, Gemma2Config, Gemma3Config, GlmConfig, Glm4Config, GPT2Config, GPT2Config,
+GPTBigCodeConfig, GPTNeoConfig, GPTNeoXConfig, GPTJConfig, HeliumConfig,
+IBertConfig, JambaConfig, JetMoeConfig, LayoutLMConfig, LayoutLMv2Config, LayoutLMv3Config, LEDConfig, LiltConfig, LlamaConfig, LongformerConfig,
+LukeConfig, MarkupLMConfig, MBartConfig, MegaConfig, MegatronBertConfig,
+MiniMaxConfig, MistralConfig, MixtralConfig, MobileBertConfig,
+ModernBertConfig, ModernBertDecoderConfig, MPNetConfig, MptConfig, MraConfig,
+MT5Config, MvpConfig, NemotronConfig, NezhaConfig, NystromformerConfig, OpenLlamaConfig, OpenAIGPTConfig, OPTConfig, PerceiverConfig, PersimmonConfig, PhiConfig, Phi3Config, PhimoeConfig, PLBartConfig, QDQBertConfig, Qwen2Config,
+Qwen2MoeConfig, Qwen3Config, Qwen3MoeConfig, ReformerConfig, RemBertConfig,
+RobertaConfig, RobertaPreLayerNormConfig, RoCBertConfig, RoFormerConfig,
+SmolLM3Config, SqueezeBertConfig, StableLmConfig, Starcoder2Config, T5Config, T5GemmaConfig, TapasConfig, TransfoXLConfig, UMT5Config, XLMCon...
+Hmm, I might have forgotten to download build. I don’t know Kaggle…
# Online
+# Build a wheel from GitHub (avoid sdists)
+git clone https://github.com/huggingface/transformers
+cd transformers
+python -m pip install -U build
+python -m build --wheel -o ../wheels
+cd ..
+python -m pip download --only-binary=:all: -d wheelhouse \
+ build setuptools wheel packaging pyproject_hooks setuptools-scm
+# Offline
+WH=/kaggle/input/<your-dataset>/wheels
+pip install --no-index --find-links=""$WH"" \
+ build setuptools wheel packaging pyproject_hooks
+Hi, I have a dataset where the text has a label that is a standardized code. The each code has a title describing the code. The data is in a pandas df called jobs_data
\ndata = {\n ""text"": jobs_data.JobTitle.to_list(),\n ""label"": jobs_data.soc2010.to_list(),\n}\nfeatures = {\n ""text"": Value(""string""),\n ""label"": ClassLabel(names=soc2010.code.to_list()),\n}\n\njobs_ds = Dataset.from_dict(data,features=Features(features))\nI would like to include a codes to title dictionary/function to make it easier to convert from a label → code → title
\nIs this possible?
\nThank you
If metadata alone is sufficient, using the DatasetInfo class is probably the quickest option.
from datasets import DatasetInfo\n\ndata = {\n ""text"": jobs_data.JobTitle.to_list(),\n ""label"": jobs_data.soc2010.to_list(),\n}\n\nfeatures = {\n ""text"": Value(""string""),\n ""label"": ClassLabel(names=soc2010.code.to_list()),\n}\n\ncode2title = ""codes to convert from a label → code → title""\n\ninfo = DatasetInfo(\n description=""Jobs dataset with SOC‐2010 codes"",\n metadata={""code2title"": code2title}\n)\n\njobs_ds = Dataset.from_dict(data, features=Features(features), info=info)\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-09-04T20:41:28.087Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 165626, 'topic_slug': 'adding-metadata-to-a-dataset', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/adding-metadata-to-a-dataset/165626/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi, I have a dataset where the text has a label that is a standardized code. The each code has a title describing the code. The data is in a pandas df called jobs_data
+data = {
+ ""text"": jobs_data.JobTitle.to_list(),
+ ""label"": jobs_data.soc2010.to_list(),
+}
+features = {
+ ""text"": Value(""string""),
+ ""label"": ClassLabel(names=soc2010.code.to_list()),
+}
+
+jobs_ds = Dataset.from_dict(data,features=Features(features))
+I would like to include a codes to title dictionary/function to make it easier to convert from a label → code → title
+Is this possible?
+Thank you
If metadata alone is sufficient, using the DatasetInfo class is probably the quickest option.
from datasets import DatasetInfo
+
+data = {
+ ""text"": jobs_data.JobTitle.to_list(),
+ ""label"": jobs_data.soc2010.to_list(),
+}
+
+features = {
+ ""text"": Value(""string""),
+ ""label"": ClassLabel(names=soc2010.code.to_list()),
+}
+
+code2title = ""codes to convert from a label → code → title""
+
+info = DatasetInfo(
+ description=""Jobs dataset with SOC‐2010 codes"",
+ metadata={""code2title"": code2title}
+)
+
+jobs_ds = Dataset.from_dict(data, features=Features(features), info=info)
+I’m new to using transformers so any help would be appreciated. I keep getting this error when I attempting to import Seq2Seq2Trainer and Seq2Seq2TrainingArguments:
\nImportError: cannot import name ‘TFPreTrainedModel’ from ‘transformers’
\nI’m not sure what to do to resolve this, I’ve already checked to make sure that transformers is up to date (version 4.56.0).
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-09-03T17:53:23.637Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 13, 'reads': 3, 'readers_count': 2, 'score': 45.6, 'yours': False, 'topic_id': 168082, 'topic_slug': 'error-importing-seq2seqtrainer', 'display_username': 'Dawson', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103089, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-importing-seq2seqtrainer/168082/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241119, 'name': 'Dawson', 'username': 'dholt123', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/f6c823/{size}.png', 'created_at': '2025-09-03T18:26:28.515Z', 'cooked': 'I was able to figure out the issue. It was caused by having both TensorFlow and pyTorch installed. When both are installed, Integration_utils.py first checks to see if TensorFlow is available first and the attempts to import TFPreTrainedModel this is where the error was occurring.
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-09-03T18:26:28.515Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 168082, 'topic_slug': 'error-importing-seq2seqtrainer', 'display_username': 'Dawson', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 103089, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-importing-seq2seqtrainer/168082/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 241148, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-04T06:27:02.281Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-09-04T06:27:02.281Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 168082, 'topic_slug': 'error-importing-seq2seqtrainer', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/error-importing-seq2seqtrainer/168082/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m new to using transformers so any help would be appreciated. I keep getting this error when I attempting to import Seq2Seq2Trainer and Seq2Seq2TrainingArguments:
+ImportError: cannot import name ‘TFPreTrainedModel’ from ‘transformers’
+I’m not sure what to do to resolve this, I’ve already checked to make sure that transformers is up to date (version 4.56.0).
","I was able to figure out the issue. It was caused by having both TensorFlow and pyTorch installed. When both are installed, Integration_utils.py first checks to see if TensorFlow is available first and the attempts to import TFPreTrainedModel this is where the error was occurring.
" +Batch generation Llama 3 Instruct | Tokenizer has no padding token,https://discuss.huggingface.co/t/batch-generation-llama-3-instruct-tokenizer-has-no-padding-token/168043,168043,9,2025-09-02 20:07:06.418000+00:00,"[{'id': 241024, 'name': 'Samir Char', 'username': 'samirchar', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/c2a13f/{size}.png', 'created_at': '2025-09-02T20:07:06.509Z', 'cooked': 'Hello everyone,
\nWhat is the best way of using a model like Llama 3.1 ( meta-llama/Llama-3.1-8B-Instruct · Hugging Face ) with AutoModel, AutoTokenizer, and template chat (I can’t use pipelines for my use case) for batch generation and eventually also using DDP.
\nThis works for a single conversation:
\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_id = ""meta-llama/Llama-3.1-8B-Instruct""\n\ntokenizer = AutoTokenizer.from_pretrained(model_id)\nmodel = AutoModelForCausalLM.from_pretrained(\n model_id, torch_dtype=torch.bfloat16, device_map=""auto""\n)\n\nmessages = [\n {\n ""role"": ""system"",\n ""content"": ""You are a pirate chatbot who always responds in pirate speak!"",\n },\n {""role"": ""user"", ""content"": ""Who are you?""},\n]\n\ninput_ids = tokenizer.apply_chat_template(\n messages, add_generation_prompt=True, return_tensors=""pt""\n).to(model.device)\n\nterminators = [\n tokenizer.eos_token_id,\n tokenizer.convert_tokens_to_ids(""<|eot_id|>""),\n]\n\noutputs = model.generate(\n input_ids,\n max_new_tokens=256,\n eos_token_id=terminators,\n do_sample=True,\n temperature=0.6,\n top_p=0.9,\n)\n\nresponse = outputs[0][input_ids.shape[-1] :]\nprint(tokenizer.decode(response, skip_special_tokens=True))\n\nFor multiple conversations and batch decoding, do I just need to apply the chat template with padding = True? When I try that, I get the error “Asking to pad but the tokenizer does not have a padding token”
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-02T20:44:24.769Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 5, 'readers_count': 4, 'score': 61.0, 'yours': False, 'topic_id': 168043, 'topic_slug': 'batch-generation-llama-3-instruct-tokenizer-has-no-padding-token', 'display_username': 'Samir Char', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct', 'internal': False, 'reflection': False, 'title': 'meta-llama/Llama-3.1-8B-Instruct · Hugging Face', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 80944, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/batch-generation-llama-3-instruct-tokenizer-has-no-padding-token/168043/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 241029, 'name': 'Samir Char', 'username': 'samirchar', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/c2a13f/{size}.png', 'created_at': '2025-09-02T20:43:55.582Z', 'cooked': 'Actually, could this be the solution?
\nSet padding to left
\nSet pad token to eos token
\nIn generate set pad token id to eos token id
\nUse tokenizer.batch_decode
\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_id = ""meta-llama/Llama-3.1-8B-Instruct""\n\ntokenizer = AutoTokenizer.from_pretrained(model_id, padding_side=""left"")\ntokenizer.pad_token = tokenizer.eos_token\ntokenizer.pad_token_id = tokenizer.eos_token_id\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_id,\n torch_dtype=torch.bfloat16,\n device_map=""auto"",\n)\n\nmessages = [\n [\n {\n ""role"": ""system"",\n ""content"": ""You are a pirate chatbot who always responds in pirate speak!"",\n },\n {""role"": ""user"", ""content"": ""Who are you?""},\n ],\n [\n {\n ""role"": ""system"",\n ""content"": ""You are a pirate chatbot who always responds in pirate speak!"",\n },\n {""role"": ""user"", ""content"": ""How old are you?""},\n ],\n]\n\ninput_ids = tokenizer.apply_chat_template(\n messages, add_generation_prompt=True, return_tensors=""pt"", padding=True\n).to(model.device)\n\nterminators = [\n tokenizer.eos_token_id,\n tokenizer.convert_tokens_to_ids(""<|eot_id|>""),\n]\n\noutputs = model.generate(\n input_ids,\n max_new_tokens=256,\n eos_token_id=terminators,\n do_sample=True,\n temperature=0.6,\n top_p=0.9,\n pad_token_id=tokenizer.eos_token_id,\n)\ntokenizer.batch_decode(outputs, skip_special_tokens=True)\n\nI think that’s correct. If anything else to add, maybe return_dict=True or something.
from transformers import AutoModelForCausalLM, AutoTokenizer\nimport torch\n\nmodel_id = ""meta-llama/Llama-3.1-8B-Instruct""\n\ntokenizer = AutoTokenizer.from_pretrained(model_id, padding_side=""left"")\ntokenizer.pad_token = tokenizer.eos_token\ntokenizer.pad_token_id = tokenizer.eos_token_id # inference-safe\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_id,\n torch_dtype=torch.bfloat16,\n device_map=""auto"",\n)\n\nmessages = [\n [\n {""role"": ""system"", ""content"": ""You are a pirate chatbot who always responds in pirate speak!""},\n {""role"": ""user"", ""content"": ""Who are you?""},\n ],\n [\n {""role"": ""system"", ""content"": ""You are a pirate chatbot who always responds in pirate speak!""},\n {""role"": ""user"", ""content"": ""How old are you?""},\n ],\n]\n\n# Return a BatchEncoding with input_ids **and** attention_mask, already padded on the left\ninputs = tokenizer.apply_chat_template(\n messages,\n add_generation_prompt=True,\n tokenize=True, # explicit\n return_tensors=""pt"",\n return_dict=True, # crucial for batched generate\n padding=True,\n).to(model.device)\n\nterminators = [tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids(""<|eot_id|>"")]\n\noutputs = model.generate(\n **inputs, # pass dict, not a single tensor\n max_new_tokens=256,\n do_sample=True,\n temperature=0.6,\n top_p=0.9,\n eos_token_id=terminators, # stop on EOS or EOT\n pad_token_id=tokenizer.eos_token_id,\n)\n\n# Drop the prompt, then decode the new tokens only\nnew_tokens = outputs[:, inputs[""input_ids""].shape[1]:]\ntexts = tokenizer.batch_decode(new_tokens, skip_special_tokens=True)\nThat’s awesome, thank you!
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-09-03T11:04:36.350Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 168043, 'topic_slug': 'batch-generation-llama-3-instruct-tokenizer-has-no-padding-token', 'display_username': 'Samir Char', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 80944, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/batch-generation-llama-3-instruct-tokenizer-has-no-padding-token/168043/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 241134, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-03T23:05:14.080Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-09-03T23:05:14.080Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 168043, 'topic_slug': 'batch-generation-llama-3-instruct-tokenizer-has-no-padding-token', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/batch-generation-llama-3-instruct-tokenizer-has-no-padding-token/168043/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone,
+What is the best way of using a model like Llama 3.1 ( meta-llama/Llama-3.1-8B-Instruct · Hugging Face ) with AutoModel, AutoTokenizer, and template chat (I can’t use pipelines for my use case) for batch generation and eventually also using DDP.
+This works for a single conversation:
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_id = ""meta-llama/Llama-3.1-8B-Instruct""
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+ model_id, torch_dtype=torch.bfloat16, device_map=""auto""
+)
+
+messages = [
+ {
+ ""role"": ""system"",
+ ""content"": ""You are a pirate chatbot who always responds in pirate speak!"",
+ },
+ {""role"": ""user"", ""content"": ""Who are you?""},
+]
+
+input_ids = tokenizer.apply_chat_template(
+ messages, add_generation_prompt=True, return_tensors=""pt""
+).to(model.device)
+
+terminators = [
+ tokenizer.eos_token_id,
+ tokenizer.convert_tokens_to_ids(""<|eot_id|>""),
+]
+
+outputs = model.generate(
+ input_ids,
+ max_new_tokens=256,
+ eos_token_id=terminators,
+ do_sample=True,
+ temperature=0.6,
+ top_p=0.9,
+)
+
+response = outputs[0][input_ids.shape[-1] :]
+print(tokenizer.decode(response, skip_special_tokens=True))
+
+For multiple conversations and batch decoding, do I just need to apply the chat template with padding = True? When I try that, I get the error “Asking to pad but the tokenizer does not have a padding token”
","I think that’s correct. If anything else to add, maybe return_dict=True or something.
from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
+
+model_id = ""meta-llama/Llama-3.1-8B-Instruct""
+
+tokenizer = AutoTokenizer.from_pretrained(model_id, padding_side=""left"")
+tokenizer.pad_token = tokenizer.eos_token
+tokenizer.pad_token_id = tokenizer.eos_token_id # inference-safe
+
+model = AutoModelForCausalLM.from_pretrained(
+ model_id,
+ torch_dtype=torch.bfloat16,
+ device_map=""auto"",
+)
+
+messages = [
+ [
+ {""role"": ""system"", ""content"": ""You are a pirate chatbot who always responds in pirate speak!""},
+ {""role"": ""user"", ""content"": ""Who are you?""},
+ ],
+ [
+ {""role"": ""system"", ""content"": ""You are a pirate chatbot who always responds in pirate speak!""},
+ {""role"": ""user"", ""content"": ""How old are you?""},
+ ],
+]
+
+# Return a BatchEncoding with input_ids **and** attention_mask, already padded on the left
+inputs = tokenizer.apply_chat_template(
+ messages,
+ add_generation_prompt=True,
+ tokenize=True, # explicit
+ return_tensors=""pt"",
+ return_dict=True, # crucial for batched generate
+ padding=True,
+).to(model.device)
+
+terminators = [tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids(""<|eot_id|>"")]
+
+outputs = model.generate(
+ **inputs, # pass dict, not a single tensor
+ max_new_tokens=256,
+ do_sample=True,
+ temperature=0.6,
+ top_p=0.9,
+ eos_token_id=terminators, # stop on EOS or EOT
+ pad_token_id=tokenizer.eos_token_id,
+)
+
+# Drop the prompt, then decode the new tokens only
+new_tokens = outputs[:, inputs[""input_ids""].shape[1]:]
+texts = tokenizer.batch_decode(new_tokens, skip_special_tokens=True)
+I preprocessed and uploaded the entirety of the gilkeyio/librispeech-alignments dataset, which is huge. However, I set the wrong dataset._info.features for one column. Now, the key_value_metadata.0.valueof every parquet file in my dataset has ""feats"": {""shape"": [null, 80], ""dtype"": ""float32"", ""_type"": ""Array2D""}when I want it to be ""feats"": {""shape"": [null, 39], ""dtype"": ""float32"", ""_type"": ""Array2D""}. Changing the README metadata doesn’t solve the problem, as I get the following error loading the dataset:
ValueError: cannot reshape array of size 8931 into shape (229,80).
How can I change the parquet metadata without processing the whole dataset once again ?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-08T14:17:33.634Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 6, 'readers_count': 5, 'score': 71.2, 'yours': False, 'topic_id': 166127, 'topic_slug': 'change-metadata-of-parquet-files', 'display_username': 'Alice Mabille', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91713, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/change-metadata-of-parquet-files/166127/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 237367, 'name': 'Sylvain Lesage', 'username': 'severo', 'avatar_template': '/user_avatar/discuss.huggingface.co/severo/{size}/27449_2.png', 'created_at': '2025-08-08T15:30:15.316Z', 'cooked': 'cc @lhoestq might know
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-08T15:30:15.316Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 166127, 'topic_slug': 'change-metadata-of-parquet-files', 'display_username': 'Sylvain Lesage', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 2900, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/change-metadata-of-parquet-files/166127/2', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240993, 'name': 'Quentin Lhoest', 'username': 'lhoestq', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png', 'created_at': '2025-09-02T10:27:16.354Z', 'cooked': 'I think you have to reprocess the data unfortunately
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-09-02T10:27:16.354Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 166127, 'topic_slug': 'change-metadata-of-parquet-files', 'display_username': 'Quentin Lhoest', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 76, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/change-metadata-of-parquet-files/166127/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 241031, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-09-02T22:27:19.321Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-09-02T22:27:19.321Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 166127, 'topic_slug': 'change-metadata-of-parquet-files', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/change-metadata-of-parquet-files/166127/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I preprocessed and uploaded the entirety of the gilkeyio/librispeech-alignments dataset, which is huge. However, I set the wrong dataset._info.features for one column. Now, the key_value_metadata.0.valueof every parquet file in my dataset has ""feats"": {""shape"": [null, 80], ""dtype"": ""float32"", ""_type"": ""Array2D""}when I want it to be ""feats"": {""shape"": [null, 39], ""dtype"": ""float32"", ""_type"": ""Array2D""}. Changing the README metadata doesn’t solve the problem, as I get the following error loading the dataset:
ValueError: cannot reshape array of size 8931 into shape (229,80).
How can I change the parquet metadata without processing the whole dataset once again ?
",I think you have to reprocess the data unfortunately
+Can I use LoRA with jhu-clsp/ettin-encoder-1b?,https://discuss.huggingface.co/t/can-i-use-lora-with-jhu-clsp-ettin-encoder-1b/167903,167903,5,2025-08-29 14:49:48.934000+00:00,"[{'id': 240628, 'name': 'Alex', 'username': 'SuperBowser', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9f8e36/{size}.png', 'created_at': '2025-08-29T14:49:49.002Z', 'cooked': 'It looks like jhu-clsp/ettin-encoder-1b does not have any proj layers. Is it possible to use LoRA with this model:
from transformers import AutoModelForSequenceClassification\nmodel_name = ‘jhu-clsp/ettin-encoder-1b’\nmodel = AutoModelForSequenceClassification.from_pretrained(model_name)\nfor parent_name, module in model.named_modules():\n for child_name, child in module.named_children():\n if ‘proj’ in child_name:\n print(child_name)\n print(“_________”)\nThis code returned nothing.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-29T14:49:49.002Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 5, 'readers_count': 4, 'score': 41.0, 'yours': False, 'topic_id': 167903, 'topic_slug': 'can-i-use-lora-with-jhu-clsp-ettin-encoder-1b', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-use-lora-with-jhu-clsp-ettin-encoder-1b/167903/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 240648, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-30T00:29:33.998Z', 'cooked': 'It seems that for ModernBERT-based models, the target_modules names aren’t proj*. You can apparently also automatically select the target_modules using =""all-linear"".
""target_modules"": [\n ""Wqkv"",\n ""Wi"",\n ""Wo""\n ],\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-09-02T14:59:52.226Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167903, 'topic_slug': 'can-i-use-lora-with-jhu-clsp-ettin-encoder-1b', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/can-i-use-lora-with-jhu-clsp-ettin-encoder-1b/167903/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","It looks like jhu-clsp/ettin-encoder-1b does not have any proj layers. Is it possible to use LoRA with this model:
from transformers import AutoModelForSequenceClassification
+model_name = ‘jhu-clsp/ettin-encoder-1b’
+model = AutoModelForSequenceClassification.from_pretrained(model_name)
+for parent_name, module in model.named_modules():
+ for child_name, child in module.named_children():
+ if ‘proj’ in child_name:
+ print(child_name)
+ print(“_________”)
+This code returned nothing.
","It seems that for ModernBERT-based models, the target_modules names aren’t proj*. You can apparently also automatically select the target_modules using =""all-linear"".
""target_modules"": [
+ ""Wqkv"",
+ ""Wi"",
+ ""Wo""
+ ],
+Hi. I finetuned mistralai/Mistral-Small-24B-Base-2501 on a dataset and now I’m trying to run inference for it. I’m using AutoModelForCausalLM.from_pretrained to load it but getting this error: Could not find MistralForCausalLM neither in transformers. I’m running the latest version of transformers 4.56.0. What might be the reason? Installing transformers from source according to this post support for MistralForCausalLM · Issue #26458 · huggingface/transformers · GitHub didn’t fix it.
Hmm, maybe it’s missing dependencies or something…?
\nI don’t think the class itself is actually missing…
pip install -U mistral_common sentencepiece\nimport transformers, sys\nprint(""transformers"", transformers.__version__)\ntry:\n from transformers.models.mistral.modeling_mistral import MistralForCausalLM\n print(""MistralForCausalLM OK"")\nexcept Exception as e:\n print(""MistralForCausalLM FAIL:"", e, file=sys.stderr)\n@John6666 getting this when I run that code snippet
\n``
\nMistralForCausalLM FAIL: partially initialized module ‘torchvision’ has no attribute ‘extension’ (most likely due to a circular import)
\n```
Judging just by the error, it’s probably a version mismatch between torch and torchvision.
pip install torchvision==x.xx.x\n@John6666 thanks! yes, aligning the versions helped
I have fine-tuned the model and now running into this run-time error while loading it:
\nRuntimeError: Error(s) in loading state_dict for Embedding:
\nsize mismatch for weight: copying a param with shape torch.Size([0]) from checkpoint, the shape in current model is torch.Size([131072, 5120]). Any idea what might be causing this?
Based on the error message, I’d guess it’s either trying to load the PEFT adapter as a whole model weight or the model weights are corrupted…
\n@John6666 could this be because of deepspeed? when I do len(tokenizer) it prints 131072.
\n\ncould this be because of deepspeed
\n
I think very likely…
\nWhen saving fails in DeepSpeed, it appears an empty tensor is saved instead.
@John6666 I’m using ""stage3_gather_16bit_weights_on_model_save"": true as suggested here. Not sure what else is causing this.
This may also occur when using BF16 or when using older version of PEFT.
\npip install -U peft\n@John6666 using model.save_16bit_model() to save the model insread of save_pretrained() fixed this!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 12, 'post_type': 3, 'posts_count': 12, 'updated_at': '2025-09-01T21:09:24.800Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 167978, 'topic_slug': 'could-not-find-mistralforcausallm-in-transformers', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/could-not-find-mistralforcausallm-in-transformers/167978/12', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi. I finetuned mistralai/Mistral-Small-24B-Base-2501 on a dataset and now I’m trying to run inference for it. I’m using AutoModelForCausalLM.from_pretrained to load it but getting this error: Could not find MistralForCausalLM neither in transformers. I’m running the latest version of transformers 4.56.0. What might be the reason? Installing transformers from source according to this post support for MistralForCausalLM · Issue #26458 · huggingface/transformers · GitHub didn’t fix it.
Judging just by the error, it’s probably a version mismatch between torch and torchvision.
pip install torchvision==x.xx.x
+Hi,
\nSome of my Gradio spaces that were working previously are no longer functioning. The first issue seems to be related to the Debian 13 update: my Gradio spaces were likely initially deployed with Debian 12.
\nAfter trying the workaround suggested by john6666, one of my older spaces restarted, but it now gets stuck with a different Python error.
\n\n\nFor another space deployed with Docker, I modified the Dockerfile to specify the Debian and Python versions:
\nFROM python:3.11-slim-bookworm\n# Instead of: FROM python:3.11-slim\nThis change was intended to use Python 3.11 with Debian 12 (Bookworm), as the default python:3.11-slim now uses Debian 13 (Trixie).
\nHowever, it initially returned an error:
E: Package \'libgl1-mesa-glx\' has no installation candidate\nAfter fixing the package error, the space no longer shows that issue, but it gets stuck during the build stage after:
\nBuilding wheel for llama-cpp-python (pyproject.toml): started\nIt get in TimeOut.
\nThe same issue occurs in a third space that was working today until I changed its name (which triggered a rebuild). Now, it also gets stuck at the same build stage.
\nFor my older spaces deployed automatically with Gradio, it would be ideal if, during a rebuild, the versions of the OS, Python, Gradio, and other essential dependencies remained the same as those used during the initial deployment. This would help avoid failures during restarts or rebuilds.
\nNote: I know that versions can be specified in requirements.txt (though not the base OS container).
For my Hugging Face Spaces that were automatically deployed for Gradio, is there a way to find out the versions of the OS, Python, and the main packages/dependencies used? This would allow me to specify or lock those versions by simply editing the requirements.txt file.
Is there a solution to stay on, for example, Debian 12 with Python 3.10 during a rebuild for spaces deployed without a Dockerfile?
\nRegarding the current error:
\nBuilding wheel for llama-cpp-python (pyproject.toml): started\nDoes specifying a version of llama-cpp-python that can be downloaded like other libraries (without needing to build a wheel) seem like the only solution?
Thank you for your feedback!
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-29T17:28:00.115Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 36, 'reads': 5, 'readers_count': 4, 'score': 141.0, 'yours': False, 'topic_id': 167908, 'topic_slug': 'broken-space-after-debian13-update-and-llama-cpp-python-update', 'display_username': 'MisterAI', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/error-get-error-when-deploy-space/166612/28', 'internal': True, 'reflection': False, 'title': '[ERROR] Get error when deploy space', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64568, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/broken-space-after-debian13-update-and-llama-cpp-python-update/167908/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 240651, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-30T01:04:49.010Z', 'cooked': '\n\n1 / 2
\n
You can specify Python versions and the additional packages to install. However, everything else must be done manually… Also, the OS is fixed in Gradio spaces.
\nimport sys, platform\nfrom importlib import metadata as md\n\nprint(""Python:"", platform.python_version(), sys.implementation.name)\nprint(""OS:"", platform.uname())\nprint(""\\n"".join(sorted(f""{d.metadata[\'Name\']}=={d.version}"" for d in md.distributions())))\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-30T01:06:22.684Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 167908, 'topic_slug': 'broken-space-after-debian13-update-and-llama-cpp-python-update', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/latest-llama-cpp-wont-build-in-spaces/166357', 'internal': True, 'reflection': False, 'title': ""Latest llama.cpp won't build in Spaces"", 'clicks': 2}, {'url': 'https://huggingface.co/docs/hub/spaces-config-reference', 'internal': False, 'reflection': False, 'title': 'Spaces Configuration Reference', 'clicks': 0}, {'url': 'https://huggingface.co/docs/hub/spaces-dependencies', 'internal': False, 'reflection': False, 'title': 'Handling Spaces Dependencies in Gradio Spaces', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/broken-space-after-debian13-update-and-llama-cpp-python-update/167908/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 240687, 'name': 'MisterAI', 'username': 'MisterAI', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/46a35a/{size}.png', 'created_at': '2025-08-30T13:14:48.891Z', 'cooked': '3
\n
hello,
\nThank you for your answer and solutions @John6666
\nAlready 2 HFSpaces up again.*
**For Memory : workaround
\n#Comment the line for llama.cpp
\n#llama-cpp-python>=0.2.0
\nimport subprocess\nimport sys, platform\nfrom importlib import metadata as md\n\n\n#Install wheel From URL (here for Python3.11 check for other python version if needed)\nsubprocess.run(""pip install https://github.com/abetlen/llama-cpp-python/releases/download/v0.3.1/llama_cpp_python-0.3.1-cp311-cp311-linux_x86_64.whl"", shell=True)\n\n#Add Log to show all versions\nprint(""Python:"", platform.python_version(), sys.implementation.name)\nprint(""OS:"", platform.uname())\nprint(""\\n"".join(sorted(f""{d.metadata[\'Name\']}=={d.version}"" for d in md.distributions())))\n\n\nimport subprocess\nimport sys, platform\nfrom importlib import metadata as md\n\n\n#Install and Compile wheel at cost of 5minutes\nsubprocess.run(""pip install -V llama_cpp_python==0.3.15"", shell=True)\n\n#Add Log to show all versions \nprint(""Python:"", platform.python_version(), sys.implementation.name)\nprint(""OS:"", platform.uname())\nprint(""\\n"".join(sorted(f""{d.metadata[\'Name\']}=={d.version}"" for d in md.distributions())))\n\nthank you.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-30T13:14:48.891Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 4, 'readers_count': 3, 'score': 35.8, 'yours': False, 'topic_id': 167908, 'topic_slug': 'broken-space-after-debian13-update-and-llama-cpp-python-update', 'display_username': 'MisterAI', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64568, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/broken-space-after-debian13-update-and-llama-cpp-python-update/167908/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240705, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-31T01:15:23.252Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-31T01:15:23.252Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167908, 'topic_slug': 'broken-space-after-debian13-update-and-llama-cpp-python-update', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/broken-space-after-debian13-update-and-llama-cpp-python-update/167908/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+Some of my Gradio spaces that were working previously are no longer functioning. The first issue seems to be related to the Debian 13 update: my Gradio spaces were likely initially deployed with Debian 12.
+After trying the workaround suggested by john6666, one of my older spaces restarted, but it now gets stuck with a different Python error.
+ + +For another space deployed with Docker, I modified the Dockerfile to specify the Debian and Python versions:
+FROM python:3.11-slim-bookworm
+# Instead of: FROM python:3.11-slim
+This change was intended to use Python 3.11 with Debian 12 (Bookworm), as the default python:3.11-slim now uses Debian 13 (Trixie).
+However, it initially returned an error:
E: Package 'libgl1-mesa-glx' has no installation candidate
+After fixing the package error, the space no longer shows that issue, but it gets stuck during the build stage after:
+Building wheel for llama-cpp-python (pyproject.toml): started
+It get in TimeOut.
+The same issue occurs in a third space that was working today until I changed its name (which triggered a rebuild). Now, it also gets stuck at the same build stage.
+For my older spaces deployed automatically with Gradio, it would be ideal if, during a rebuild, the versions of the OS, Python, Gradio, and other essential dependencies remained the same as those used during the initial deployment. This would help avoid failures during restarts or rebuilds.
+Note: I know that versions can be specified in requirements.txt (though not the base OS container).
For my Hugging Face Spaces that were automatically deployed for Gradio, is there a way to find out the versions of the OS, Python, and the main packages/dependencies used? This would allow me to specify or lock those versions by simply editing the requirements.txt file.
Is there a solution to stay on, for example, Debian 12 with Python 3.10 during a rebuild for spaces deployed without a Dockerfile?
+Regarding the current error:
+Building wheel for llama-cpp-python (pyproject.toml): started
+Does specifying a version of llama-cpp-python that can be downloaded like other libraries (without needing to build a wheel) seem like the only solution?
Thank you for your feedback!
","++1 / 2
+
You can specify Python versions and the additional packages to install. However, everything else must be done manually… Also, the OS is fixed in Gradio spaces.
+import sys, platform
+from importlib import metadata as md
+
+print(""Python:"", platform.python_version(), sys.implementation.name)
+print(""OS:"", platform.uname())
+print(""\n"".join(sorted(f""{d.metadata['Name']}=={d.version}"" for d in md.distributions())))
+++" +Which data parallel does trainer use? DP or DDP?,https://discuss.huggingface.co/t/which-data-parallel-does-trainer-use-dp-or-ddp/16021,16021,9,2022-03-24 06:03:27.073000+00:00,"[{'id': 33067, 'name': 'dr_xiami', 'username': 'xiami', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/x/dc4da7/{size}.png', 'created_at': '2022-03-24T06:03:27.154Z', 'cooked': '3
+
I try to search in the doc. But I didn’t find the answer anywhere.
\nThank you
', 'post_number': 1, 'post_type': 1, 'posts_count': 7, 'updated_at': '2022-03-24T06:03:27.154Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5299, 'reads': 205, 'readers_count': 204, 'score': 26516.0, 'yours': False, 'topic_id': 16021, 'topic_slug': 'which-data-parallel-does-trainer-use-dp-or-ddp', 'display_username': 'dr_xiami', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 3838, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/which-data-parallel-does-trainer-use-dp-or-ddp/16021/1', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 33091, 'name': 'Sylvain Gugger', 'username': 'sgugger', 'avatar_template': '/user_avatar/discuss.huggingface.co/sgugger/{size}/2291_2.png', 'created_at': '2022-03-24T12:22:07.153Z', 'cooked': 'It depends if you launch your training script with python (in which case it will use DP) or python -m torch.distributed.launch (in which case it will use DDP).
perhaps useful to you: Using Transformers with DistributedDataParallel — any examples?
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2022-08-17T15:03:18.063Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 47, 'reads': 193, 'readers_count': 192, 'score': 318.6, 'yours': False, 'topic_id': 16021, 'topic_slug': 'which-data-parallel-does-trainer-use-dp-or-ddp', 'display_username': 'Brando Miranda', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/using-transformers-with-distributeddataparallel-any-examples/10775', 'internal': True, 'reflection': False, 'title': 'Using Transformers with DistributedDataParallel — any examples?', 'clicks': 1940}, {'url': 'https://discuss.huggingface.co/t/how-to-run-an-end-to-end-example-of-distributed-data-parallel-with-hugging-faces-trainer-api-ideally-on-a-single-node-multiple-gpus/21750', 'internal': True, 'reflection': True, 'title': ""How to run an end to end example of distributed data parallel with hugging face's trainer api (ideally on a single node multiple gpus)?"", 'clicks': 16}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 3664, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/which-data-parallel-does-trainer-use-dp-or-ddp/16021/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 3}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240653, 'name': 'Rylan Schaeffer', 'username': 'RylanSchaeffer', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/6f9a4e/{size}.png', 'created_at': '2025-08-30T01:34:06.356Z', 'cooked': 'I know this is a bit of an old thread, but I have a follow up question. I’m creating a Trainer() , evaluating, training and evaluating again. Here’s a snippet of my code:
```
\ntrainer = Trainer(
\nmodel=model,
\nprocessing_class=tokenizer,
\nargs=pretraining_config,
\ntrain_dataset=train_dataset,
\neval_dataset=eval_dataset,
\ndata_collator=data_collator,
\n)
logging.info(“Evaluating before training…”)
\neval_metrics_before = trainer.evaluate()
\nwandb.log({f""eval_before/{k}"": v for k, v in eval_metrics_before.items()})
\npprint.pprint(eval_metrics_before)
logging.info(“Beginning training…”)
\ntrainer.train()
logging.info(“Finished training. Beginning final evaluation…”)
\neval_metrics_after = trainer.evaluate()
\nwandb.log({f""eval_after/{k}"": v for k, v in eval_metrics_after.items()})
\npprint.pprint(eval_metrics_after)
\n```
When I run with two GPUs and a model small enough to fit on each, I noticed while the job is running that evaluating appears to use data parallelism over the two visible GPUs, but does not for training. Do you know what might cause that or how to fix it?
', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-30T01:34:56.436Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 16021, 'topic_slug': 'which-data-parallel-does-trainer-use-dp-or-ddp', 'display_username': 'Rylan Schaeffer', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4145, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/which-data-parallel-does-trainer-use-dp-or-ddp/16021/4', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240654, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-30T02:42:00.790Z', 'cooked': 'Hmm… Have you tried launching it via accelerate or torchrun?
# single node, 2 GPUs\ntorchrun --nproc_per_node=2 train.py\n# or\naccelerate launch --num_processes=2 train.py\nYeah, I would’ve thought that launching with python would use DP and thus would only use 1 available GPU. And that’s partially correct: train() indeed only uses 1 GPU, but evaluate() uses 2 GPUs. Hence my confusion…
I see. When running distributed training, if you launch it as a single process, evaluate sometimes behaves differently from the Trainer part…Since DP itself seems quite fragile, using DDP is probably the simpler approach…
I try to search in the doc. But I didn’t find the answer anywhere.
+Thank you
",It depends if you launch your training script with python (in which case it will use DP) or python -m torch.distributed.launch (in which case it will use DDP).
Hello, I wasn’t sure if I should use the category transformers, datasets, or tokenizers for this, but wanted to post some benchmark times for training a GPT style tokenizer on a 10s of GB text dataset because they seem slower than my expectation (which could be totally off). The pre-processing sequences step took ~ 3 hours on a modern 12 core AMD CPU.
\nHere is the script I used
\nimport datasets \nfrom transformers import AutoTokenizer \n \ndef batch_iterator(dataset, batch_size=1_000): \n for batch in dataset.iter(batch_size=batch_size): \n yield batch[""text""] \n \nif __name__ == ""__main__"": \n \n ds_id = ""gabrielaltay/pubtator-central-bigbio-kb-2022-12-18"" \n clone_from_name = ""gpt2"" \n vocab_size = 32_768 \n \n clone_from_tokenizer = AutoTokenizer.from_pretrained(clone_from_name) \n ds_train = datasets.load_dataset(ds_id, split=""train"") \n \n tokenizer = clone_from_tokenizer.train_new_from_iterator( \n batch_iterator(ds_train), \n vocab_size=vocab_size, \n ) \n \n tokenizer.save_pretrained(""pubtator-gpt2-v32k-tokenizer"")\nand here is the output,
\npython train_tokenizer.py\nNone of PyTorch, TensorFlow >= 2.0, or Flax have been found. Models won\'t be available and only tokenizers, configuration and file/data utilities can be used.\nUsing custom data configuration gabrielaltay--pubtator-central-bigbio-kb-2022-12-18-51c5a8a315ecf808\nFound cached dataset parquet (/home/galtay/.cache/huggingface/datasets/gabrielaltay___parquet/gabrielaltay--pubtator-central-bigbio-kb-2022-12-18-51c5a8a315ecf808/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec)\n[02:55:09] Pre-processing sequences █████████████████████████████ 0 / 0\n[00:00:07] Tokenize words █████████████████████████████ 6828518 / 6828518\n[00:00:13] Count pairs █████████████████████████████ 6828518 / 6828518\n[00:00:48] Compute merges █████████████████████████████ 32511 / 32511\nThe train split of the dataset is ~100GB but the text is duplicated in another column with markup so I estimate about 50GB in the “text” column. I think this should be doable at “training a tokenizer on english wikipedia speeds” within a factor of 10 or so (I was thinking minutes not hours). Can anyone see where I’m making a mistake or if my time estimates are just totally off?
\nI’m using,
\ndatasets 2.8.0
\ntransformers 4.25.1
and this is the dataset on the hub gabrielaltay/pubtator-central-bigbio-kb-2022-12-18 · Datasets at Hugging Face
\nthanks,
\n-G
UPDATE: attempting to isolate dataset iteration speed with
\nimport datasets \nfrom tqdm import tqdm \nimport datasets \n \ndef batch_iterator(dataset, batch_size=1_000): \n for batch in dataset.iter(batch_size=batch_size): \n yield batch[""text""] \n \nif __name__ == ""__main__"": \n \n ds_id = ""gabrielaltay/pubtator-central-bigbio-kb-2022-12-18"" \n ds_train = datasets.load_dataset(ds_id, split=""train"") \n for batch in tqdm(batch_iterator(ds_train)): \n x = 1 \nand getting,
\n700it [02:10, 5.18it/s]\nleading me to believe the bottleneck is dataset iteration speed
\n(33M samples) / (batch size 1000) / (6 it/s) = 5500 s ~ 90 minutes
Problem Solved! (thanks to @lhoestq)
\nTurns out the slow iteration speed was b/c of all the extra columns in the dataset besides the “text” column. Running with just the text column in the dataset gave 40x speedup ,
\nold\n700it [02:10, 5.18it/s]\n\nnew\n13435it [00:32, 228.80it/s]\nimport datasets \nfrom transformers import AutoTokenizer \n \ndef batch_iterator(dataset, batch_size=1_000): \n for batch in dataset.iter(batch_size=batch_size): \n yield batch[""text""] \n \nif __name__ == ""__main__"": \n \n ds_id = ""gabrielaltay/pubtator-central-bigbio-kb-2022-12-18"" \n clone_from_name = ""gpt2"" \n vocab_size = 32_768 \n \n clone_from_tokenizer = AutoTokenizer.from_pretrained(clone_from_name) \n ds_train = datasets.load_dataset(ds_id, split=""train"") \n # remove non text columns\n ds_train = ds_train.remove_columns([ \n col for col in ds_train.column_names if col != ""text"" \n ]) \n \n tokenizer = clone_from_tokenizer.train_new_from_iterator( \n batch_iterator(ds_train), \n vocab_size=vocab_size, \n ) \n \n tokenizer.save_pretrained(""pubtator-gpt2-v32k-tokenizer"") \nI also have the issue of slow training speed with the tokenizer on smaller datasets. Upon investigation, it became clear that the tokenizer only utilizes 1 CPU core, and batching or not batching doesn’t affect its speed. What do you think is the solution to this problem?
', 'post_number': 3, 'post_type': 1, 'posts_count': 9, 'updated_at': '2024-03-04T10:07:12.613Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 39, 'readers_count': 38, 'score': 102.8, 'yours': False, 'topic_id': 29125, 'topic_slug': 'speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset', 'display_username': 'Mahdi Masoon', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 42772, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset/29125/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 128372, 'name': 'Haris Jabbar', 'username': 'maveriq', 'avatar_template': '/user_avatar/discuss.huggingface.co/maveriq/{size}/27075_2.png', 'created_at': '2024-05-01T10:10:39.032Z', 'cooked': 'I agree. The training doesn’t seem to be using all cores; and it’s still bottlenecked by the rate at which data can be read from the iterator.
\nI wonder if there is any way to improve that.
', 'post_number': 4, 'post_type': 1, 'posts_count': 9, 'updated_at': '2024-05-01T10:10:39.032Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 34, 'readers_count': 33, 'score': 46.8, 'yours': False, 'topic_id': 29125, 'topic_slug': 'speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset', 'display_username': 'Haris Jabbar', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 42772, 'username': 'MahdiMasoon', 'name': 'Mahdi Masoon', 'avatar_template': '/user_avatar/discuss.huggingface.co/mahdimasoon/{size}/34330_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 1294, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset/29125/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 141049, 'name': 'Karandeep Singh', 'username': 'kdcyberdude', 'avatar_template': '/user_avatar/discuss.huggingface.co/kdcyberdude/{size}/27478_2.png', 'created_at': '2024-07-01T16:06:22.056Z', 'cooked': 'Hi @gabrielaltay, I am facing the same issue… I am currently training a BPE tokenizer for the Panjabi language on a 50 GB text corpus. However, I am encountering an “Out of Memory” (OOM) issue even when using a 1TB RAM instance. Can you help me understand the reason behind this and provide any references or suggestions to train this model more efficiently?
\nfrom datasets import load_from_disk, load_dataset\nfrom transformers import AutoTokenizer\n\nds = load_dataset(\'kdcyberdude/Vichaar\', num_proc=8, cache_dir=\'./gemma_data_cache\')[\'train\']\nprint(ds)\ntokenizer = AutoTokenizer.from_pretrained(""openchat/openchat-3.5-0106-gemma"")\n\ndef batch_iterator(batch_size=1000):\n for i in range(0, len(ds), batch_size):\n yield ds[i : i + batch_size][""text""]\n\nnew_tokenizer = tokenizer.train_new_from_iterator( batch_iterator(), vocab_size=32_000, length=len(ds))\nnew_tokenizer.save_pretrained(""./gemma-32k-pa-tokenizer"")\nI have also tried this using a data loader, the Pre-processing sequences steps keep on iterating even after len(ds) and memory keeps increasing. The iteration goes 7*len(ds) until it hits OOM. Not sure when it will stop. Same as this issue and issue
\nclass TextDataset(torch.utils.data.Dataset):\n def __init__(self, ds, batch_size):\n self.batch_size = batch_size\n self.ds = ds\n\n def __len__(self):\n return len(self.ds)\n\n def __getitem__(self, idx):\n batch = self.ds[idx:idx + self.batch_size][\'text\']\n return batch\n\ndataset = TextDataset(ds, batch_size=1024)\ndataloader = torch.utils.data.DataLoader(dataset, batch_size=None)\n\nnew_tokenizer = tokenizer.train_new_from_iterator( dataloader, vocab_size=32_000, length=len(ds))\nI also tried debugging the code to understand which part is consuming this much RAM but I am not able to get into this train_from_iterator function in tokenization_utils_fast.py. I am speculating this could be calling executable/binary code that may be running in Rust.
Any help or pointers would be greatly appreciated!
\n
That is indeed weird, I’ll investigate as it should be using threads
', 'post_number': 6, 'post_type': 1, 'posts_count': 9, 'updated_at': '2024-07-16T08:49:51.872Z', 'reply_count': 1, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 26, 'reads': 24, 'readers_count': 23, 'score': 139.8, 'yours': False, 'topic_id': 29125, 'topic_slug': 'speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset', 'display_username': 'Arthur Zucker', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 36632, 'username': 'kdcyberdude', 'name': 'Karandeep Singh', 'avatar_template': '/user_avatar/discuss.huggingface.co/kdcyberdude/{size}/27478_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 7005, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset/29125/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 146420, 'name': 'Arthur Zucker', 'username': 'ArthurZ', 'avatar_template': '/user_avatar/discuss.huggingface.co/arthurz/{size}/26972_2.png', 'created_at': '2024-07-26T10:16:45.611Z', 'cooked': 'Fast encode by ArthurZucker · Pull Request #1560 · huggingface/tokenizers · GitHub should help! There are issue with parallelization
', 'post_number': 7, 'post_type': 1, 'posts_count': 9, 'updated_at': '2024-07-26T10:16:45.611Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 22, 'readers_count': 21, 'score': 34.4, 'yours': False, 'topic_id': 29125, 'topic_slug': 'speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset', 'display_username': 'Arthur Zucker', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/tokenizers/pull/1560', 'internal': False, 'reflection': False, 'title': 'Fast encode by ArthurZucker · Pull Request #1560 · huggingface/tokenizers · GitHub', 'clicks': 94}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 7005, 'username': 'ArthurZ', 'name': 'Arthur Zucker', 'avatar_template': '/user_avatar/discuss.huggingface.co/arthurz/{size}/26972_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 7005, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/speed-issues-using-tokenizer-train-new-from-iterator-on-50gb-dataset/29125/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 169291, 'name': 'Leon Lee', 'username': 'Leon-Leee', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/l/ecb155/{size}.png', 'created_at': '2024-11-11T04:16:50.428Z', 'cooked': 'Hi, I encountered the same problem as @kdcyberdude did. I used a host with 1.5TB memory and trained a 64k-vocab tokenizer on a 25GB text corpus using hf tokenizer. It ran slower and slower and broke down during merging.
\nCould anyone tell me how to avoid this?
Same here. The tokenizer trainer seems to be using only 1 core.
\nAlso, I want to stream the dataset so when dealing with huge dataset it won’t OOM.
I am pretty new so correct me if I am doing it wrong:
\n# I know wikitext isn\'t large but in case I need to deal with large dataset\ndataset_dict = load_dataset(""wikitext"", ""wikitext-103-raw-v1"", streaming=True)\nsplits = [dataset_dict[k] for k in dataset_dict] # use all splits\ndataset = interleave_datasets(splits, stopping_strategy=""all_exhausted"")\n\ndef batch_iterator(dataset, batch_size=1_000): \n for batch in dataset.iter(batch_size=batch_size): \n yield batch[""text""]\n\ntokenizer = ByteLevelBPETokenizer()\ntokenizer.train_from_iterator(\n batch_iterator(dataset),\n vocab_size=30000,\n min_frequency=2,\n special_tokens=[""<pad>"", ""<unk>"", ""<bos>"", ""<eos>""],\n show_progress=True,\n)\nHello, I wasn’t sure if I should use the category transformers, datasets, or tokenizers for this, but wanted to post some benchmark times for training a GPT style tokenizer on a 10s of GB text dataset because they seem slower than my expectation (which could be totally off). The pre-processing sequences step took ~ 3 hours on a modern 12 core AMD CPU.
+Here is the script I used
+import datasets
+from transformers import AutoTokenizer
+
+def batch_iterator(dataset, batch_size=1_000):
+ for batch in dataset.iter(batch_size=batch_size):
+ yield batch[""text""]
+
+if __name__ == ""__main__"":
+
+ ds_id = ""gabrielaltay/pubtator-central-bigbio-kb-2022-12-18""
+ clone_from_name = ""gpt2""
+ vocab_size = 32_768
+
+ clone_from_tokenizer = AutoTokenizer.from_pretrained(clone_from_name)
+ ds_train = datasets.load_dataset(ds_id, split=""train"")
+
+ tokenizer = clone_from_tokenizer.train_new_from_iterator(
+ batch_iterator(ds_train),
+ vocab_size=vocab_size,
+ )
+
+ tokenizer.save_pretrained(""pubtator-gpt2-v32k-tokenizer"")
+and here is the output,
+python train_tokenizer.py
+None of PyTorch, TensorFlow >= 2.0, or Flax have been found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
+Using custom data configuration gabrielaltay--pubtator-central-bigbio-kb-2022-12-18-51c5a8a315ecf808
+Found cached dataset parquet (/home/galtay/.cache/huggingface/datasets/gabrielaltay___parquet/gabrielaltay--pubtator-central-bigbio-kb-2022-12-18-51c5a8a315ecf808/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec)
+[02:55:09] Pre-processing sequences █████████████████████████████ 0 / 0
+[00:00:07] Tokenize words █████████████████████████████ 6828518 / 6828518
+[00:00:13] Count pairs █████████████████████████████ 6828518 / 6828518
+[00:00:48] Compute merges █████████████████████████████ 32511 / 32511
+The train split of the dataset is ~100GB but the text is duplicated in another column with markup so I estimate about 50GB in the “text” column. I think this should be doable at “training a tokenizer on english wikipedia speeds” within a factor of 10 or so (I was thinking minutes not hours). Can anyone see where I’m making a mistake or if my time estimates are just totally off?
+I’m using,
+datasets 2.8.0
+transformers 4.25.1
and this is the dataset on the hub gabrielaltay/pubtator-central-bigbio-kb-2022-12-18 · Datasets at Hugging Face
+thanks,
+-G
UPDATE: attempting to isolate dataset iteration speed with
+import datasets
+from tqdm import tqdm
+import datasets
+
+def batch_iterator(dataset, batch_size=1_000):
+ for batch in dataset.iter(batch_size=batch_size):
+ yield batch[""text""]
+
+if __name__ == ""__main__"":
+
+ ds_id = ""gabrielaltay/pubtator-central-bigbio-kb-2022-12-18""
+ ds_train = datasets.load_dataset(ds_id, split=""train"")
+ for batch in tqdm(batch_iterator(ds_train)):
+ x = 1
+and getting,
+700it [02:10, 5.18it/s]
+leading me to believe the bottleneck is dataset iteration speed
+(33M samples) / (batch size 1000) / (6 it/s) = 5500 s ~ 90 minutes
Problem Solved! (thanks to @lhoestq)
+Turns out the slow iteration speed was b/c of all the extra columns in the dataset besides the “text” column. Running with just the text column in the dataset gave 40x speedup ,
+old
+700it [02:10, 5.18it/s]
+
+new
+13435it [00:32, 228.80it/s]
+import datasets
+from transformers import AutoTokenizer
+
+def batch_iterator(dataset, batch_size=1_000):
+ for batch in dataset.iter(batch_size=batch_size):
+ yield batch[""text""]
+
+if __name__ == ""__main__"":
+
+ ds_id = ""gabrielaltay/pubtator-central-bigbio-kb-2022-12-18""
+ clone_from_name = ""gpt2""
+ vocab_size = 32_768
+
+ clone_from_tokenizer = AutoTokenizer.from_pretrained(clone_from_name)
+ ds_train = datasets.load_dataset(ds_id, split=""train"")
+ # remove non text columns
+ ds_train = ds_train.remove_columns([
+ col for col in ds_train.column_names if col != ""text""
+ ])
+
+ tokenizer = clone_from_tokenizer.train_new_from_iterator(
+ batch_iterator(ds_train),
+ vocab_size=vocab_size,
+ )
+
+ tokenizer.save_pretrained(""pubtator-gpt2-v32k-tokenizer"")
+Hi. I’m trying to fine-tune mistralai/Mistral-Small-24B-Base-2501 using deepspeed and consistently getting the overflow error. When I use bf16 and fp32,I don’t see the overflow issue but the training loss is Nan. When I switch to fp16 the training loss is correct but it throws the overflow error. How can I fix this? This works fine with smaller models. Using lr=1e-7.
My df_config.json:
{\n ""train_micro_batch_size_per_gpu"": 1,\n ""gradient_accumulation_steps"": 8,\n ""zero_optimization"": {\n ""stage"": 2\n },\n ""zero_allow_untested_optimizer"": true,\n ""fp16"": {\n ""enabled"": true,\n ""loss_scale"": 0,\n ""initial_scale_power"": 32,\n ""loss_scale_window"": 1000,\n ""hysteresis"": 2,\n ""min_loss_scale"": 1\n },\n ""gradient_clipping"": 1.0,\n ""wall_clock_breakdown"": false\n}\nUsing deepspeed 0.17.2 and transformers 4.42.4.
If the GPU supports bfloat16, it’s probably better to use bfloat16. Regarding NaN issues, SDPA seems to be the culprit in many cases. Try attn_implementation=""eager"".
@John6666 loading the model in bfloat16 and then using bf16=true in deepspeed seems to solve this issue for now!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-28T16:51:04.376Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167833, 'topic_slug': 'gradient-overflow-issue-while-using-deepspeed', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/gradient-overflow-issue-while-using-deepspeed/167833/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi. I’m trying to fine-tune mistralai/Mistral-Small-24B-Base-2501 using deepspeed and consistently getting the overflow error. When I use bf16 and fp32,I don’t see the overflow issue but the training loss is Nan. When I switch to fp16 the training loss is correct but it throws the overflow error. How can I fix this? This works fine with smaller models. Using lr=1e-7.
My df_config.json:
{
+ ""train_micro_batch_size_per_gpu"": 1,
+ ""gradient_accumulation_steps"": 8,
+ ""zero_optimization"": {
+ ""stage"": 2
+ },
+ ""zero_allow_untested_optimizer"": true,
+ ""fp16"": {
+ ""enabled"": true,
+ ""loss_scale"": 0,
+ ""initial_scale_power"": 32,
+ ""loss_scale_window"": 1000,
+ ""hysteresis"": 2,
+ ""min_loss_scale"": 1
+ },
+ ""gradient_clipping"": 1.0,
+ ""wall_clock_breakdown"": false
+}
+Using deepspeed 0.17.2 and transformers 4.42.4.
If the GPU supports bfloat16, it’s probably better to use bfloat16. Regarding NaN issues, SDPA seems to be the culprit in many cases. Try attn_implementation=""eager"".
Hello,
\nI am trying to put my hands on transformers (this is my first project with transformers). I decided to do a bert2bert translator, as it one of those tested in the following paper https://arxiv.org/pdf/1907.12461
\nI put my tests here Bert2Bert_translator/Bert_translator.ipynb at 0fb904c480df2a2de53f51e9b9198b65b6fcf770 · jclary-31/Bert2Bert_translator · GitHub
\nI used the EncoderDecoderModel to combine one Bert in encoder mode and another one in decoder mode. I then fine tuned the model but something is off…
\nmaybe it is because I use the wrong Bert checkpoint, maybe it is because encoder inputs are not correct (but this step should be automatic, maybe it is something else. Should I separate encoder and decoder?
I don’t know where the problem lies,
\nI tried on bigger dataset, it changes nothing. In the end my final in a translation task will still be something as ‘ [CLS] [CLS] [CLS]…’. So I think the issue is in the conception. Something I missed or understand wrong.
I checked in forums, Github, website, and found no concrete example on such translator…
\nDo you know what is wrong? It is in the code or in the conception?
\nThanks
', 'post_number': 1, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-17T22:57:32.379Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 4, 'readers_count': 3, 'score': 40.8, 'yours': False, 'topic_id': 167108, 'topic_slug': 'bert2bert-translator', 'display_username': 'jean clary', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://arxiv.org/pdf/1907.12461', 'internal': False, 'reflection': False, 'clicks': 1}, {'url': 'https://github.com/jclary-31/Bert2Bert_translator/blob/0fb904c480df2a2de53f51e9b9198b65b6fcf770/Bert_translator.ipynb', 'internal': False, 'reflection': False, 'title': 'Bert2Bert_translator/Bert_translator.ipynb at 0fb904c480df2a2de53f51e9b9198b65b6fcf770 · jclary-31/Bert2Bert_translator · GitHub', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101949, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/bert2bert-translator/167108/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239023, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-18T01:40:59.887Z', 'cooked': 'There seem to be several known cases. I tried having AI write some demo code.
\nimport torch\nfrom transformers import (\n BertTokenizerFast, BertConfig, BertLMHeadModel, BertModel,\n AutoModel, EncoderDecoderModel, AutoTokenizer, AutoModelForSeq2SeqLM\n)\n\ntorch.manual_seed(0)\nenc = dec = ""bert-base-uncased""\ntok_src = BertTokenizerFast.from_pretrained(enc)\ntok_tgt = BertTokenizerFast.from_pretrained(dec)\n\n# ---------- WRONG_1: BOS loop risk (labels include BOS + manual decoder_input_ids)\ndec_cfg = BertConfig.from_pretrained(dec, is_decoder=True, add_cross_attention=True)\nbad_train = EncoderDecoderModel(\n encoder=AutoModel.from_pretrained(enc),\n decoder=BertLMHeadModel.from_pretrained(dec, config=dec_cfg),\n)\nX = tok_src([""i like tea""], return_tensors=""pt"", padding=True, truncation=True)\nY = tok_tgt([""j\'aime le thé""], return_tensors=""pt"", padding=True, truncation=True) # has [CLS]\nlabels = Y.input_ids.clone(); labels[labels == tok_tgt.pad_token_id] = -100\n_ = bad_train(input_ids=X[""input_ids""], attention_mask=X[""attention_mask""],\n decoder_input_ids=Y.input_ids, labels=labels) # ❌\ngen = bad_train.generate(\n X[""input_ids""], attention_mask=X[""attention_mask""], max_new_tokens=8,\n decoder_start_token_id=tok_tgt.cls_token_id, eos_token_id=tok_tgt.sep_token_id, pad_token_id=tok_tgt.pad_token_id\n)\nprint(""WRONG_1 gen ids:"", gen[0][:8].tolist())\n\n# ---------- WRONG_2: decoder lacks LM head / cross-attn\nplain_decoder = BertModel.from_pretrained(dec) # ❌\nbroken = EncoderDecoderModel(encoder=AutoModel.from_pretrained(enc), decoder=plain_decoder)\ntry:\n lbl2 = tok_tgt([""les chats sont mignons""], return_tensors=""pt"",\n padding=True, truncation=True, add_special_tokens=False).input_ids\n lbl2[lbl2 == tok_tgt.pad_token_id] = -100\n _ = broken(input_ids=X[""input_ids""], attention_mask=X[""attention_mask""], labels=lbl2)\n print(""WRONG_2 ran (decoder misconfigured)"")\nexcept Exception as e:\n print(""WRONG_2 error:"", type(e).__name__)\n\n# ---------- CORRECT: set decoder_start_token_id ON CONFIG before forward\ndec_cfg_ok = BertConfig.from_pretrained(dec, is_decoder=True, add_cross_attention=True)\ngood = EncoderDecoderModel(\n encoder=AutoModel.from_pretrained(enc),\n decoder=BertLMHeadModel.from_pretrained(dec, config=dec_cfg_ok),\n)\n# Required for loss computation (right-shift uses this)\ngood.config.decoder_start_token_id = tok_tgt.cls_token_id\ngood.config.eos_token_id = tok_tgt.sep_token_id\ngood.config.pad_token_id = tok_tgt.pad_token_id\ngood.config.vocab_size = good.config.decoder.vocab_size\ngood.config.tie_encoder_decoder = False\n\nX2 = tok_src([""cats are cute"", ""i like tea""], return_tensors=""pt"", padding=True, truncation=True)\nY2 = tok_tgt([""les chats sont mignons"", ""j\'aime le thé""], return_tensors=""pt"",\n padding=True, truncation=True, add_special_tokens=False) # no [CLS]\nlabels2 = Y2.input_ids.clone(); labels2[labels2 == tok_tgt.pad_token_id] = -100\n_ = good(input_ids=X2[""input_ids""], attention_mask=X2[""attention_mask""], labels=labels2) # ✅ no error\n\ngen2 = good.generate(\n X2[""input_ids""], attention_mask=X2[""attention_mask""],\n num_beams=4, max_new_tokens=24, no_repeat_ngram_size=3, early_stopping=True,\n decoder_start_token_id=tok_tgt.cls_token_id, eos_token_id=tok_tgt.sep_token_id, pad_token_id=tok_tgt.pad_token_id\n)\nprint(""CORRECT gen:"", [tok_tgt.decode(g, skip_special_tokens=True) for g in gen2])\n\n# ---------- CHECK: known-good BERT2BERT\nname = ""google/bert2bert_L-24_wmt_en_de""\ntok_g = AutoTokenizer.from_pretrained(name, pad_token=""<pad>"", bos_token=""<s>"", eos_token=""</s>"")\nmdl_g = AutoModelForSeq2SeqLM.from_pretrained(name)\nids = tok_g(""Would you like a coffee?"", return_tensors=""pt"", add_special_tokens=False).input_ids\nprint(""CHECK gen:"", tok_g.decode(mdl_g.generate(ids, num_beams=4, max_new_tokens=32)[0], skip_special_tokens=True))\n\n#WRONG_1 gen ids: [101, 6730, 6730, 6730, 6730, 6730, 6730, 6730]\n#WRONG_2 error: ValueError\n#CORRECT gen: [\'played rule rule rule rules rule rule play rule play play rule rule pass rule play pass rule rule win rule rule flow rule\', \'the. and and and pass pass pass rule rule rule pass pass be rule rule be rule pass rule pass be pass pass\']\n#CHECK gen: Haben Sie Lust auf einen Kaffee?\nhello
\nI made a small and quick test code following your advices Bert2Bert_translator/bert2bert_quicktest.ipynb at main · jclary-31/Bert2Bert_translator · GitHub
\nSo,
\nwhen no_repeat_ngram is in the parameters, some word will be generated, without this parameters the same word is repeated again and again. It is like the ‘#CORRECT gen: [\'played rule rule rule rules rule rule’ in your last answer.
\n\nIn my main code, where i test fine tuning, if I don’t use the parameter norepeat_ngram, the text generated remain ‘[CLS] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] …’
\nIf I use the parameters norepeat_ngram=3, the text generated is
\n[CLS] [PAD] [PAD] [PAD], [PAD] [PAD] of [PAD] [PAD] and [PAD] [PAD]esian [PAD] [PAD] lucas [PAD] [PAD]chfield [PAD]
So I think there is still head attention issues. Do you you know how to fix it? Should I update the Bert_translator.ipynb on github so you can see it?
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-24T18:23:41.161Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 167108, 'topic_slug': 'bert2bert-translator', 'display_username': 'jean clary', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/jclary-31/Bert2Bert_translator/blob/main/bert2bert_quicktest.ipynb', 'internal': False, 'reflection': False, 'title': 'Bert2Bert_translator/bert2bert_quicktest.ipynb at main · jclary-31/Bert2Bert_translator · GitHub', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101949, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/bert2bert-translator/167108/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240148, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-25T00:00:15.736Z', 'cooked': 'The above solution is just to suppress PAD tokens…
\nWhen actually implementing this, you will need to perform actual training and use a tokenizer that supports both languages.
# pip install -U transformers datasets\nimport random, math\nimport torch\nfrom torch.utils.data import DataLoader\nfrom torch.optim import AdamW\nfrom datasets import load_dataset\nfrom transformers import (\n AutoTokenizer, AutoModel, BertConfig, BertLMHeadModel, EncoderDecoderModel\n)\n\n# ---- config\nSEED = 0\nSRC_CKPT = ""bert-base-uncased"" # encoder (EN)\nTGT_CKPT = ""bert-base-multilingual-cased"" # decoder (FR-capable)\nMAX_SRC_LEN = 96\nMAX_TGT_LEN = 96\nBATCH_SIZE = 8\nEPOCHS = 10 # raise to 20–30 if not overfitting\nLR = 5e-5\n\nrandom.seed(SEED)\ntorch.manual_seed(SEED)\ndevice = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")\n\n# ---- tokenizers\ntok_src = AutoTokenizer.from_pretrained(SRC_CKPT)\ntok_tgt = AutoTokenizer.from_pretrained(TGT_CKPT)\nPAD_ID = tok_tgt.pad_token_id\nEOS_ID = tok_tgt.sep_token_id\nBOS_ID = tok_tgt.cls_token_id\n\n# ---- model: BERT encoder + BERT LM-head decoder with cross-attn\ndec_cfg = BertConfig.from_pretrained(TGT_CKPT, is_decoder=True, add_cross_attention=True)\nmodel = EncoderDecoderModel(\n encoder=AutoModel.from_pretrained(SRC_CKPT),\n decoder=BertLMHeadModel.from_pretrained(TGT_CKPT, config=dec_cfg),\n).to(device)\n# required special ids for training (right-shift) and decode\nmodel.config.decoder_start_token_id = BOS_ID\nmodel.config.eos_token_id = EOS_ID\nmodel.config.pad_token_id = PAD_ID\nmodel.config.tie_encoder_decoder = False\nmodel.config.vocab_size = model.config.decoder.vocab_size\n\n# ---- tiny EN–FR set: take 100 pairs from OPUS Books\n# notes: you can replace this with your own parallel lists\nds = load_dataset(""Helsinki-NLP/opus_books"", ""en-fr"", split=""train"") # ~1M pairs\npairs = [(ex[""translation""][""en""], ex[""translation""][""fr""]) for ex in ds.select(range(2000))]\nrandom.shuffle(pairs)\npairs = pairs[:100] # exactly 100\nsrc_list, tgt_list = zip(*pairs)\n\n# ---- helpers\ndef build_batch(src_texts, tgt_texts):\n # source\n X = tok_src(\n list(src_texts), padding=True, truncation=True, max_length=MAX_SRC_LEN, return_tensors=""pt""\n )\n # target labels: NO BOS; append EOS; mask PAD with -100\n Y = tok_tgt(\n list(tgt_texts), padding=""max_length"", truncation=True, max_length=MAX_TGT_LEN,\n add_special_tokens=False, return_tensors=""pt""\n )[""input_ids""]\n # append EOS before padding if room\n Y_fixed = torch.full_like(Y, PAD_ID)\n for i in range(Y.size(0)):\n toks = [t for t in Y[i].tolist() if t != PAD_ID]\n if len(toks) < MAX_TGT_LEN:\n toks = toks + [EOS_ID]\n toks = toks[:MAX_TGT_LEN]\n Y_fixed[i, :len(toks)] = torch.tensor(toks, dtype=Y_fixed.dtype)\n labels = Y_fixed.clone()\n labels[labels == PAD_ID] = -100\n\n return {k: v.to(device) for k, v in X.items()}, labels.to(device)\n\ndef collate(batch):\n s, t = zip(*batch)\n return build_batch(s, t)\n\n# simple Dataset wrapper\nclass Pairs(torch.utils.data.Dataset):\n def __init__(self, srcs, tgts):\n self.s = list(srcs); self.t = list(tgts)\n def __len__(self): return len(self.s)\n def __getitem__(self, i): return self.s[i], self.t[i]\n\ntrain_dl = DataLoader(Pairs(src_list, tgt_list), batch_size=BATCH_SIZE, shuffle=True, collate_fn=collate)\n\n@torch.inference_mode()\ndef translate_samples(texts, n=5):\n X = tok_src(list(texts[:n]), return_tensors=""pt"", padding=True, truncation=True, max_length=MAX_SRC_LEN).to(device)\n out = model.generate(\n X[""input_ids""], attention_mask=X[""attention_mask""],\n num_beams=4, max_new_tokens=64, early_stopping=True,\n decoder_start_token_id=BOS_ID, eos_token_id=EOS_ID, pad_token_id=PAD_ID,\n bad_words_ids=[[PAD_ID]], # block PAD\n repetition_penalty=1.1, # mild\n no_repeat_ngram_size=3 # optional hygiene\n )\n return [tok_tgt.decode(o, skip_special_tokens=True) for o in out]\n\ndef show_before_after(k=5):\n print(""\\n--- BEFORE ---"")\n preds_before = translate_samples(src_list, n=k)\n for i in range(k):\n print(f""EN: {src_list[i]}"")\n print(f""FR_gold: {tgt_list[i]}"")\n print(f""FR_pred: {preds_before[i]}"")\n print(""-"")\n # train then test again\n model.train()\n opt = AdamW(model.parameters(), lr=LR)\n steps = 0\n for epoch in range(EPOCHS):\n for X, labels in train_dl:\n opt.zero_grad()\n out = model(input_ids=X[""input_ids""], attention_mask=X[""attention_mask""], labels=labels)\n out.loss.backward()\n torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)\n opt.step()\n steps += 1\n print(f""epoch {epoch+1}/{EPOCHS} done"")\n model.eval()\n\n print(""\\n--- AFTER ---"")\n preds_after = translate_samples(src_list, n=k)\n for i in range(k):\n print(f""EN: {src_list[i]}"")\n print(f""FR_gold: {tgt_list[i]}"")\n print(f""FR_pred: {preds_after[i]}"")\n print(""-"")\n\nif __name__ == ""__main__"":\n print(f""device: {device}"")\n show_before_after(k=5)\n\n""""""\n--- BEFORE ---\nEN: As for me, I found myself obliged, the first time for months, to face alone a long Thursday evening - with the clear feeling that the old carriage had borne away my youth forever.\nFR_gold: Quant à moi, je me trouvai, pour la première fois depuis de longs mois, seul en face d’une longue soirée de jeudi – avec l’impression que, dans cette vieille voiture, mon adolescence venait de s’en aller pour toujours.\nFR_pred: ##iiilililiililiiliiliilingingiingiingiingingingingiiliiliingiingiiliiliigingingillingingighingiingingiingiiliingingiiliingiigiingiingieningingioviingiinginiingiingiiingiingighinginginingingiigingi\n-\nEN: No one asked him who Booby was.\nFR_gold: Personne ne lui demanda qui était Ganache.\nFR_pred: a a a - - - a a A A A a a ad ad ad Ad Ad Ad ad ad a a, a a ae ae ae a A a A,, A A, - -,,, a,,. - - an an an,, an an - - A A - - 1 -\n-\nEN: M. Seurel\'s here .. .\'\nFR_gold: M. Seurel est là…\nFR_pred: ##ggg22233322443344423243234377799988877889979773378789786779777688\n-\nEN: After the ball where everything was charming but feverish and mad, where he had himself so madly chased the tall Pierrot, Meaulnes found that he had dropped into the most peaceful happiness on earth.\nFR_gold: Après cette fête où tout était charmant, mais fiévreux et fou, où lui-même avait si follement poursuivi le grand pierrot, Meaulnes se trouvait là plongé dans le bonheur le plus calme du monde.\nFR_pred: ##iiilililiiiiliilililiiliiliigiigiigiiliiliiliingiingiingiiliilingingingiingiingiigiigingingiigiigiingiingingingiiliigiingiigingiingiigiingingiingingiigiingiiciingiingificiingiingiiciigiigiiciingi\n-\nEN: At half-past eight, just as M. Seurel was giving the signal to enter school, we arrived, quite out of breath, to line up.\nFR_gold: À huit heures et demie, à l’instant où M. Seurel allait donner le signal d’entrer, nous arrivâmes tout essoufflés pour nous mettre sur les rangs.\nFR_pred: ##jajajajanjanjanjajajanojanjanjaljanjan sal sal saljanjan sino sino sinojanjanjanojanojanojanjano sino sinojanojano sal salcolcolcolcalcalcalcolcol sal salsal sal salallallall sal sal alcolcolsalsalcolcol - - sal sal\n-\n\n--- AFTER ---\nEN: As for me, I found myself obliged, the first time for months, to face alone a long Thursday evening - with the clear feeling that the old carriage had borne away my youth forever.\nFR_gold: Quant à moi, je me trouvai, pour la première fois depuis de longs mois, seul en face d’une longue soirée de jeudi – avec l’impression que, dans cette vieille voiture, mon adolescence venait de s’en aller pour toujours.\nFR_pred: Quant à moi, je ne voulus pas pour la première fois de soi, seul en face d une longue longue aventure de longs mois.\n-\nEN: No one asked him who Booby was.\nFR_gold: Personne ne lui demanda qui était Ganache.\nFR_pred: Personne ne lui demanda qui demanda demanda qui lui demanda demanda qu il demanda Ganache.\n-\nEN: M. Seurel\'s here .. .\'\nFR_gold: M. Seurel est là…\nFR_pred: M. Seurel est là\n-\nEN: After the ball where everything was charming but feverish and mad, where he had himself so madly chased the tall Pierrot, Meaulnes found that he had dropped into the most peaceful happiness on earth.\nFR_gold: Après cette fête où tout était charmant, mais fiévreux et fou, où lui-même avait si follement poursuivi le grand pierrot, Meaulnes se trouvait là plongé dans le bonheur le plus calme du monde.\nFR_pred: Dès qu on le recommença plus le grand pierrot de sa société où lui même même même avait si beau.\n-\nEN: At half-past eight, just as M. Seurel was giving the signal to enter school, we arrived, quite out of breath, to line up.\nFR_gold: À huit heures et demie, à l’instant où M. Seurel allait donner le signal d’entrer, nous arrivâmes tout essoufflés pour nous mettre sur les rangs.\nFR_pred: À huit heures et demie à peine, nous arrivâmes tout tout essoufflés sur les rangs.\n-\n""""""\nhello John, thank you very much for your help.
\nso,
\nThanks a lot for your help, I learned a lot. For example I was not aware of the repetition_penalty, nor the no_repeat_ngram_size parameters.
\nif I may ask , why model.config.tie_encoder_decoder = False?
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-27T17:58:19.222Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 167108, 'topic_slug': 'bert2bert-translator', 'display_username': 'jean clary', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101949, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/bert2bert-translator/167108/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240469, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-27T23:53:08.081Z', 'cooked': '\n\nwhy model.config.tie_encoder_decoder = False?
\n
I thought it would be problematic if this parameter were set to True when using it across two or more models.
\n', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-27T23:53:08.081Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 167108, 'topic_slug': 'bert2bert-translator', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/main_classes/configuration#transformers.PretrainedConfig.tie_encoder_decoder', 'internal': False, 'reflection': False, 'title': 'Configuration', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/tied-weights-for-encoder-and-decoder-vocab-matrix-hard-coded-in-t5/37572', 'internal': True, 'reflection': False, 'title': 'Tied weights for encoder and decoder vocab matrix hard coded in T5?', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/bert2bert-translator/167108/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240511, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-28T11:53:20.716Z', 'cooked': 'tie_encoder_decoder (
\nbool, optional, defaults toFalse) — Whether all encoder weights should be tied to their equivalent decoder weights. This requires the encoder and decoder model to have the exact same parameter names.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-08-28T11:53:20.716Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167108, 'topic_slug': 'bert2bert-translator', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/bert2bert-translator/167108/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+I am trying to put my hands on transformers (this is my first project with transformers). I decided to do a bert2bert translator, as it one of those tested in the following paper https://arxiv.org/pdf/1907.12461
+I put my tests here Bert2Bert_translator/Bert_translator.ipynb at 0fb904c480df2a2de53f51e9b9198b65b6fcf770 · jclary-31/Bert2Bert_translator · GitHub
+I used the EncoderDecoderModel to combine one Bert in encoder mode and another one in decoder mode. I then fine tuned the model but something is off…
+maybe it is because I use the wrong Bert checkpoint, maybe it is because encoder inputs are not correct (but this step should be automatic, maybe it is something else. Should I separate encoder and decoder?
I don’t know where the problem lies,
+I tried on bigger dataset, it changes nothing. In the end my final in a translation task will still be something as ‘ [CLS] [CLS] [CLS]…’. So I think the issue is in the conception. Something I missed or understand wrong.
I checked in forums, Github, website, and found no concrete example on such translator…
+Do you know what is wrong? It is in the code or in the conception?
+Thanks
","The above solution is just to suppress PAD tokens…
+When actually implementing this, you will need to perform actual training and use a tokenizer that supports both languages.
# pip install -U transformers datasets
+import random, math
+import torch
+from torch.utils.data import DataLoader
+from torch.optim import AdamW
+from datasets import load_dataset
+from transformers import (
+ AutoTokenizer, AutoModel, BertConfig, BertLMHeadModel, EncoderDecoderModel
+)
+
+# ---- config
+SEED = 0
+SRC_CKPT = ""bert-base-uncased"" # encoder (EN)
+TGT_CKPT = ""bert-base-multilingual-cased"" # decoder (FR-capable)
+MAX_SRC_LEN = 96
+MAX_TGT_LEN = 96
+BATCH_SIZE = 8
+EPOCHS = 10 # raise to 20–30 if not overfitting
+LR = 5e-5
+
+random.seed(SEED)
+torch.manual_seed(SEED)
+device = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")
+
+# ---- tokenizers
+tok_src = AutoTokenizer.from_pretrained(SRC_CKPT)
+tok_tgt = AutoTokenizer.from_pretrained(TGT_CKPT)
+PAD_ID = tok_tgt.pad_token_id
+EOS_ID = tok_tgt.sep_token_id
+BOS_ID = tok_tgt.cls_token_id
+
+# ---- model: BERT encoder + BERT LM-head decoder with cross-attn
+dec_cfg = BertConfig.from_pretrained(TGT_CKPT, is_decoder=True, add_cross_attention=True)
+model = EncoderDecoderModel(
+ encoder=AutoModel.from_pretrained(SRC_CKPT),
+ decoder=BertLMHeadModel.from_pretrained(TGT_CKPT, config=dec_cfg),
+).to(device)
+# required special ids for training (right-shift) and decode
+model.config.decoder_start_token_id = BOS_ID
+model.config.eos_token_id = EOS_ID
+model.config.pad_token_id = PAD_ID
+model.config.tie_encoder_decoder = False
+model.config.vocab_size = model.config.decoder.vocab_size
+
+# ---- tiny EN–FR set: take 100 pairs from OPUS Books
+# notes: you can replace this with your own parallel lists
+ds = load_dataset(""Helsinki-NLP/opus_books"", ""en-fr"", split=""train"") # ~1M pairs
+pairs = [(ex[""translation""][""en""], ex[""translation""][""fr""]) for ex in ds.select(range(2000))]
+random.shuffle(pairs)
+pairs = pairs[:100] # exactly 100
+src_list, tgt_list = zip(*pairs)
+
+# ---- helpers
+def build_batch(src_texts, tgt_texts):
+ # source
+ X = tok_src(
+ list(src_texts), padding=True, truncation=True, max_length=MAX_SRC_LEN, return_tensors=""pt""
+ )
+ # target labels: NO BOS; append EOS; mask PAD with -100
+ Y = tok_tgt(
+ list(tgt_texts), padding=""max_length"", truncation=True, max_length=MAX_TGT_LEN,
+ add_special_tokens=False, return_tensors=""pt""
+ )[""input_ids""]
+ # append EOS before padding if room
+ Y_fixed = torch.full_like(Y, PAD_ID)
+ for i in range(Y.size(0)):
+ toks = [t for t in Y[i].tolist() if t != PAD_ID]
+ if len(toks) < MAX_TGT_LEN:
+ toks = toks + [EOS_ID]
+ toks = toks[:MAX_TGT_LEN]
+ Y_fixed[i, :len(toks)] = torch.tensor(toks, dtype=Y_fixed.dtype)
+ labels = Y_fixed.clone()
+ labels[labels == PAD_ID] = -100
+
+ return {k: v.to(device) for k, v in X.items()}, labels.to(device)
+
+def collate(batch):
+ s, t = zip(*batch)
+ return build_batch(s, t)
+
+# simple Dataset wrapper
+class Pairs(torch.utils.data.Dataset):
+ def __init__(self, srcs, tgts):
+ self.s = list(srcs); self.t = list(tgts)
+ def __len__(self): return len(self.s)
+ def __getitem__(self, i): return self.s[i], self.t[i]
+
+train_dl = DataLoader(Pairs(src_list, tgt_list), batch_size=BATCH_SIZE, shuffle=True, collate_fn=collate)
+
+@torch.inference_mode()
+def translate_samples(texts, n=5):
+ X = tok_src(list(texts[:n]), return_tensors=""pt"", padding=True, truncation=True, max_length=MAX_SRC_LEN).to(device)
+ out = model.generate(
+ X[""input_ids""], attention_mask=X[""attention_mask""],
+ num_beams=4, max_new_tokens=64, early_stopping=True,
+ decoder_start_token_id=BOS_ID, eos_token_id=EOS_ID, pad_token_id=PAD_ID,
+ bad_words_ids=[[PAD_ID]], # block PAD
+ repetition_penalty=1.1, # mild
+ no_repeat_ngram_size=3 # optional hygiene
+ )
+ return [tok_tgt.decode(o, skip_special_tokens=True) for o in out]
+
+def show_before_after(k=5):
+ print(""\n--- BEFORE ---"")
+ preds_before = translate_samples(src_list, n=k)
+ for i in range(k):
+ print(f""EN: {src_list[i]}"")
+ print(f""FR_gold: {tgt_list[i]}"")
+ print(f""FR_pred: {preds_before[i]}"")
+ print(""-"")
+ # train then test again
+ model.train()
+ opt = AdamW(model.parameters(), lr=LR)
+ steps = 0
+ for epoch in range(EPOCHS):
+ for X, labels in train_dl:
+ opt.zero_grad()
+ out = model(input_ids=X[""input_ids""], attention_mask=X[""attention_mask""], labels=labels)
+ out.loss.backward()
+ torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
+ opt.step()
+ steps += 1
+ print(f""epoch {epoch+1}/{EPOCHS} done"")
+ model.eval()
+
+ print(""\n--- AFTER ---"")
+ preds_after = translate_samples(src_list, n=k)
+ for i in range(k):
+ print(f""EN: {src_list[i]}"")
+ print(f""FR_gold: {tgt_list[i]}"")
+ print(f""FR_pred: {preds_after[i]}"")
+ print(""-"")
+
+if __name__ == ""__main__"":
+ print(f""device: {device}"")
+ show_before_after(k=5)
+
+""""""
+--- BEFORE ---
+EN: As for me, I found myself obliged, the first time for months, to face alone a long Thursday evening - with the clear feeling that the old carriage had borne away my youth forever.
+FR_gold: Quant à moi, je me trouvai, pour la première fois depuis de longs mois, seul en face d’une longue soirée de jeudi – avec l’impression que, dans cette vieille voiture, mon adolescence venait de s’en aller pour toujours.
+FR_pred: ##iiilililiililiiliiliilingingiingiingiingingingingiiliiliingiingiiliiliigingingillingingighingiingingiingiiliingingiiliingiigiingiingieningingioviingiinginiingiingiiingiingighinginginingingiigingi
+-
+EN: No one asked him who Booby was.
+FR_gold: Personne ne lui demanda qui était Ganache.
+FR_pred: a a a - - - a a A A A a a ad ad ad Ad Ad Ad ad ad a a, a a ae ae ae a A a A,, A A, - -,,, a,,. - - an an an,, an an - - A A - - 1 -
+-
+EN: M. Seurel's here .. .'
+FR_gold: M. Seurel est là…
+FR_pred: ##ggg22233322443344423243234377799988877889979773378789786779777688
+-
+EN: After the ball where everything was charming but feverish and mad, where he had himself so madly chased the tall Pierrot, Meaulnes found that he had dropped into the most peaceful happiness on earth.
+FR_gold: Après cette fête où tout était charmant, mais fiévreux et fou, où lui-même avait si follement poursuivi le grand pierrot, Meaulnes se trouvait là plongé dans le bonheur le plus calme du monde.
+FR_pred: ##iiilililiiiiliilililiiliiliigiigiigiiliiliiliingiingiingiiliilingingingiingiingiigiigingingiigiigiingiingingingiiliigiingiigingiingiigiingingiingingiigiingiiciingiingificiingiingiiciigiigiiciingi
+-
+EN: At half-past eight, just as M. Seurel was giving the signal to enter school, we arrived, quite out of breath, to line up.
+FR_gold: À huit heures et demie, à l’instant où M. Seurel allait donner le signal d’entrer, nous arrivâmes tout essoufflés pour nous mettre sur les rangs.
+FR_pred: ##jajajajanjanjanjajajanojanjanjaljanjan sal sal saljanjan sino sino sinojanjanjanojanojanojanjano sino sinojanojano sal salcolcolcolcalcalcalcolcol sal salsal sal salallallall sal sal alcolcolsalsalcolcol - - sal sal
+-
+
+--- AFTER ---
+EN: As for me, I found myself obliged, the first time for months, to face alone a long Thursday evening - with the clear feeling that the old carriage had borne away my youth forever.
+FR_gold: Quant à moi, je me trouvai, pour la première fois depuis de longs mois, seul en face d’une longue soirée de jeudi – avec l’impression que, dans cette vieille voiture, mon adolescence venait de s’en aller pour toujours.
+FR_pred: Quant à moi, je ne voulus pas pour la première fois de soi, seul en face d une longue longue aventure de longs mois.
+-
+EN: No one asked him who Booby was.
+FR_gold: Personne ne lui demanda qui était Ganache.
+FR_pred: Personne ne lui demanda qui demanda demanda qui lui demanda demanda qu il demanda Ganache.
+-
+EN: M. Seurel's here .. .'
+FR_gold: M. Seurel est là…
+FR_pred: M. Seurel est là
+-
+EN: After the ball where everything was charming but feverish and mad, where he had himself so madly chased the tall Pierrot, Meaulnes found that he had dropped into the most peaceful happiness on earth.
+FR_gold: Après cette fête où tout était charmant, mais fiévreux et fou, où lui-même avait si follement poursuivi le grand pierrot, Meaulnes se trouvait là plongé dans le bonheur le plus calme du monde.
+FR_pred: Dès qu on le recommença plus le grand pierrot de sa société où lui même même même avait si beau.
+-
+EN: At half-past eight, just as M. Seurel was giving the signal to enter school, we arrived, quite out of breath, to line up.
+FR_gold: À huit heures et demie, à l’instant où M. Seurel allait donner le signal d’entrer, nous arrivâmes tout essoufflés pour nous mettre sur les rangs.
+FR_pred: À huit heures et demie à peine, nous arrivâmes tout tout essoufflés sur les rangs.
+-
+""""""
+>>> generator = pipeline(""text-generation"", model=""HuggingFaceTB/SmolLM2-360M"")\nconfig.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 689/689 [00:00<00:00, 415kB/s]\nmodel.safetensors: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 724M/724M [00:09<00:00, 73.1MB/s]\ngeneration_config.json: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 111/111 [00:00<00:00, 697kB/s]\ntokenizer_config.json: 3.66kB [00:00, 10.4MB/s]\nvocab.json: 801kB [00:00, 9.48MB/s]\nmerges.txt: 466kB [00:00, 36.9MB/s]\ntokenizer.json: 2.10MB [00:00, 53.9MB/s]\nspecial_tokens_map.json: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 831/831 [00:00<00:00, 1.66MB/s]\nDevice set to use mps:0\n>>> generator(""I\'m not sure if I know how to"", max_length=50, num_return_sequences=3,)\nTruncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation=True` to explicitly truncate examples to max length. Defaulting to \'longest_first\' truncation strategy. If you encode pairs of sequences (GLUE-style) with the tokenizer you can select this strategy more precisely by providing a specific strategy to `truncation`.\nSetting `pad_token_id` to `eos_token_id`:0 for open-end generation.\nBoth `max_new_tokens` (=256) and `max_length`(=50) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)\n[{\'generated_text\': ""I\'m not sure if I know how to explain this. The problem basically is that you can\'t have a value of 0 in the output. I\'m trying to do the following:\\n\\nfloat x = 2.0;\\nfloat y = 0.0;\\nfloat z = 1.0;\\nfloat z2;\\n\\nz2 = z + x*y;\\n\\nI understand that y*z should be 2.0*0.0 = 0.0, but I\'m not sure how to get the 0.0 in the z2 variable.\\n\\n## Answers\\n\\n0\\n1. If you are trying to get the 0.0 in z2, please look at the following code:\\nbool true = (z2*z2) > 0;\\n\\n// The result is 0.0\\n\\nfloat z2 = z2*z2;\\n\\n// The result is 0.0\\n\\nfloat z2 = z2*z2*z2;\\n\\n// The result is 0.0\\n\\n## Re: How to get 0 in a value in the output in a function\\n\\nThanks for the reply! I understand the problem now.\\n\\nI was trying""}, {\'generated_text\': ""I\'m not sure if I know how to do that.\\n\\nHow can I find the derivative of 1/x?\\n\\nI can\'t find the derivative of x^3\\n\\nI can\'t find the derivative of x^1/2\\n\\nI can\'t find the derivative of x^1/3\\n\\nI can\'t find the derivative of x^1/4\\n\\nI can\'t find the derivative of x^1/5\\n\\nI can\'t find the derivative of x^1/6\\n\\nI can\'t find the derivative of x^1/7\\n\\nI can\'t find the derivative of x^1/8\\n\\nI can\'t find the derivative of x^1/9\\n\\nI can\'t find the derivative of x^10\\n\\nI can\'t find the derivative of x^11\\n\\nI can\'t find the derivative of x^12\\n\\nI can\'t find the derivative of x^13\\n\\nI can\'t find the derivative of x^14\\n\\nI can\'t find the derivative of x^15\\n\\nI can\'t find the derivative of x^16\\n\\nI can\'t find the derivative of x^17\\n\\nI can\'t find the derivative of x^""}, {\'generated_text\': ""I\'m not sure if I know how to do this, but I tried to make a function that generates the 64 bit numbers and I got 128 bit numbers.\\n\\n```function rand64(digits = 128) {\\nconst digits = digits;\\nconst d = 7;\\nconst s = 2147483647;\\nconst e = -2147483648;\\nconst f = 1;\\nconst g = 2;\\nconst h = 3;\\nconst i = 4;\\n\\nconst m = 1024;\\nconst d1 = 1 << d;\\nconst d2 = 1 << d - d1;\\nconst d3 = 1 << d - d1 - d2;\\nconst d4 = 1 << d - d1 - d2 - d3;\\nconst d5 = 1 << d - d1 - d2 - d3 - d4;\\nconst d6 = 1 << d - d1 - d2 - d3 - d4 - d5;\\nconst d7 = 1 << d - d1 - d2 - d3 - d4 - d""}]\n\nIt doesn’t seem like the max_length is being honored when this is run. This is straight out of the LLM course under the “Transformers, what can they do?” section.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-27T00:53:51.147Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 13, 'reads': 7, 'readers_count': 6, 'score': 81.4, 'yours': False, 'topic_id': 167794, 'topic_slug': 'setting-max-length-does-not-limit-length-of-output', 'display_username': 'Travis Lelle', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102600, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/setting-max-length-does-not-limit-length-of-output/167794/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 240366, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-27T03:20:49.986Z', 'cooked': 'With the current Transformers library code, max_new_tokens takes precedence over max_length, so specifying max_new_tokens is the simplest approach.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-27T15:21:13.240Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 167794, 'topic_slug': 'setting-max-length-does-not-limit-length-of-output', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/setting-max-length-does-not-limit-length-of-output/167794/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]",">>> generator = pipeline(""text-generation"", model=""HuggingFaceTB/SmolLM2-360M"")
+config.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 689/689 [00:00<00:00, 415kB/s]
+model.safetensors: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 724M/724M [00:09<00:00, 73.1MB/s]
+generation_config.json: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 111/111 [00:00<00:00, 697kB/s]
+tokenizer_config.json: 3.66kB [00:00, 10.4MB/s]
+vocab.json: 801kB [00:00, 9.48MB/s]
+merges.txt: 466kB [00:00, 36.9MB/s]
+tokenizer.json: 2.10MB [00:00, 53.9MB/s]
+special_tokens_map.json: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 831/831 [00:00<00:00, 1.66MB/s]
+Device set to use mps:0
+>>> generator(""I'm not sure if I know how to"", max_length=50, num_return_sequences=3,)
+Truncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation=True` to explicitly truncate examples to max length. Defaulting to 'longest_first' truncation strategy. If you encode pairs of sequences (GLUE-style) with the tokenizer you can select this strategy more precisely by providing a specific strategy to `truncation`.
+Setting `pad_token_id` to `eos_token_id`:0 for open-end generation.
+Both `max_new_tokens` (=256) and `max_length`(=50) seem to have been set. `max_new_tokens` will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)
+[{'generated_text': ""I'm not sure if I know how to explain this. The problem basically is that you can't have a value of 0 in the output. I'm trying to do the following:\n\nfloat x = 2.0;\nfloat y = 0.0;\nfloat z = 1.0;\nfloat z2;\n\nz2 = z + x*y;\n\nI understand that y*z should be 2.0*0.0 = 0.0, but I'm not sure how to get the 0.0 in the z2 variable.\n\n## Answers\n\n0\n1. If you are trying to get the 0.0 in z2, please look at the following code:\nbool true = (z2*z2) > 0;\n\n// The result is 0.0\n\nfloat z2 = z2*z2;\n\n// The result is 0.0\n\nfloat z2 = z2*z2*z2;\n\n// The result is 0.0\n\n## Re: How to get 0 in a value in the output in a function\n\nThanks for the reply! I understand the problem now.\n\nI was trying""}, {'generated_text': ""I'm not sure if I know how to do that.\n\nHow can I find the derivative of 1/x?\n\nI can't find the derivative of x^3\n\nI can't find the derivative of x^1/2\n\nI can't find the derivative of x^1/3\n\nI can't find the derivative of x^1/4\n\nI can't find the derivative of x^1/5\n\nI can't find the derivative of x^1/6\n\nI can't find the derivative of x^1/7\n\nI can't find the derivative of x^1/8\n\nI can't find the derivative of x^1/9\n\nI can't find the derivative of x^10\n\nI can't find the derivative of x^11\n\nI can't find the derivative of x^12\n\nI can't find the derivative of x^13\n\nI can't find the derivative of x^14\n\nI can't find the derivative of x^15\n\nI can't find the derivative of x^16\n\nI can't find the derivative of x^17\n\nI can't find the derivative of x^""}, {'generated_text': ""I'm not sure if I know how to do this, but I tried to make a function that generates the 64 bit numbers and I got 128 bit numbers.\n\n```function rand64(digits = 128) {\nconst digits = digits;\nconst d = 7;\nconst s = 2147483647;\nconst e = -2147483648;\nconst f = 1;\nconst g = 2;\nconst h = 3;\nconst i = 4;\n\nconst m = 1024;\nconst d1 = 1 << d;\nconst d2 = 1 << d - d1;\nconst d3 = 1 << d - d1 - d2;\nconst d4 = 1 << d - d1 - d2 - d3;\nconst d5 = 1 << d - d1 - d2 - d3 - d4;\nconst d6 = 1 << d - d1 - d2 - d3 - d4 - d5;\nconst d7 = 1 << d - d1 - d2 - d3 - d4 - d""}]
+
+It doesn’t seem like the max_length is being honored when this is run. This is straight out of the LLM course under the “Transformers, what can they do?” section.
","With the current Transformers library code, max_new_tokens takes precedence over max_length, so specifying max_new_tokens is the simplest approach.
Hi. This looks like an issue from peft side. I’m working with mistralai/Mistral-Small-24B-Base-2501 model and trying to fine-tune it. But it throws ImportError: cannot import name ‘PreTrainedModel’ from ‘transformers’. My versions are transformers 4.55.4, tokenizers 0.21.4, peft 0.17.1. Is this a version incompatibility issue?
I downgraded transformers to 4.42.4, tokenizers to 0.19.1 and peft to 0.5.0 and it throws Exception: data did not match any variant of untagged enum ModelWrapper at line 1217944 column 3
Managed to solve this by using tokenizers-0.20.1 transformers-4.45.2 (json - Tokenizer.from_file() HUGGINFACE : Exception: data did not match any variant of untagged enum ModelWrapper - Stack Overflow)
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-27T03:01:32.882Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 26, 'reads': 14, 'readers_count': 13, 'score': 127.4, 'yours': False, 'topic_id': 167797, 'topic_slug': 'importerror-cannot-import-name-pretrainedmodel-from-transformers', 'display_username': 'Jay', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://stackoverflow.com/a/79076471', 'internal': False, 'reflection': False, 'title': 'json - Tokenizer.from_file() HUGGINFACE : Exception: data did not match any variant of untagged enum ModelWrapper - Stack Overflow', 'clicks': 63}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 16838, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/importerror-cannot-import-name-pretrainedmodel-from-transformers/167797/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 240414, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-27T15:02:11.108Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-27T15:02:11.108Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 9, 'readers_count': 8, 'score': 41.4, 'yours': False, 'topic_id': 167797, 'topic_slug': 'importerror-cannot-import-name-pretrainedmodel-from-transformers', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/importerror-cannot-import-name-pretrainedmodel-from-transformers/167797/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi. This looks like an issue from peft side. I’m working with mistralai/Mistral-Small-24B-Base-2501 model and trying to fine-tune it. But it throws ImportError: cannot import name ‘PreTrainedModel’ from ‘transformers’. My versions are transformers 4.55.4, tokenizers 0.21.4, peft 0.17.1. Is this a version incompatibility issue?
I downgraded transformers to 4.42.4, tokenizers to 0.19.1 and peft to 0.5.0 and it throws Exception: data did not match any variant of untagged enum ModelWrapper at line 1217944 column 3
Managed to solve this by using tokenizers-0.20.1 transformers-4.45.2 (json - Tokenizer.from_file() HUGGINFACE : Exception: data did not match any variant of untagged enum ModelWrapper - Stack Overflow)
" +Cannot import name ‘_resolve_process_group’ from ‘torch.distributed.distributed_c10d’,https://discuss.huggingface.co/t/cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d/167762,167762,9,2025-08-25 19:56:34.430000+00:00,"[{'id': 240239, 'name': 'Elizabeth Wainwright', 'username': 'ewainwright', 'avatar_template': '/user_avatar/discuss.huggingface.co/ewainwright/{size}/53052_2.png', 'created_at': '2025-08-25T19:56:34.479Z', 'cooked': 'I got the following error when calling the HuggingFaceLLM class:
\nFailed to import transformers.generation.utils because of the following error (look up to see its traceback): cannot import name \'_resolve_process_group\' from \'torch.distributed.distributed_c10d\'\nI looked into the source code and sure enough that function is not in there. Is this a versioning problem?
\nUpdate: I downgraded transformers to version 4.27.4 and that seemed to solve that issue but now I have a keyerror for “mistral”. Is there anyway I can solve this issue without downgrading transformers?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-25T20:47:38.847Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 24, 'reads': 3, 'readers_count': 2, 'score': 135.6, 'yours': False, 'topic_id': 167762, 'topic_slug': 'cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d', 'display_username': 'Elizabeth Wainwright', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102505, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d/167762/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 240260, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-26T00:33:05.978Z', 'cooked': 'This error seems to occur when PyTorch is far older than Transformers. It should be OK with PyTorch 2.4 or later.
\nimport torch, torch.distributed as dist\nprint(torch.__version__, \'dist?\', dist.is_available())\n# Expect: 2.4+ dist? True\nThanks this worked
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-26T12:32:16.124Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 167762, 'topic_slug': 'cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d', 'display_username': 'Elizabeth Wainwright', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102505, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d/167762/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240358, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-27T00:32:22.645Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-27T00:32:22.645Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167762, 'topic_slug': 'cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cannot-import-name-resolve-process-group-from-torch-distributed-distributed-c10d/167762/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I got the following error when calling the HuggingFaceLLM class:
+Failed to import transformers.generation.utils because of the following error (look up to see its traceback): cannot import name '_resolve_process_group' from 'torch.distributed.distributed_c10d'
+I looked into the source code and sure enough that function is not in there. Is this a versioning problem?
+Update: I downgraded transformers to version 4.27.4 and that seemed to solve that issue but now I have a keyerror for “mistral”. Is there anyway I can solve this issue without downgrading transformers?
","This error seems to occur when PyTorch is far older than Transformers. It should be OK with PyTorch 2.4 or later.
+import torch, torch.distributed as dist
+print(torch.__version__, 'dist?', dist.is_available())
+# Expect: 2.4+ dist? True
+Hello everyone!
\nI’m using a Docker Space to deploy my FastAPI application that uses multiple models, but I’ve set it to private since my project contains sensitive code. My problem is that I can’t send requests to the endpoints from anywhere outside my browser and get a 404.
Is it possible to send a token with the request to authenticate myself? If so, how should I include it in my request to make it work properly?
Thank you all in advance!
If the space is functioning properly, you should be able to access it like following.
\nYou can figure out the actual space URL yourself, also you can also find it using the GUI.
curl -X POST https://OWNER-SPACENAME.hf.space/api/predict \\\n -H ""Authorization: Bearer $HF_TOKEN"" \\\n -H ""Content-Type: application/json"" \\\n -d \'{""text"":""hello""}\'\nor
\nimport os, requests\nurl = ""https://OWNER-SPACENAME.hf.space/api/predict""\nr = requests.post(url,\n headers={""Authorization"": f""Bearer {os.getenv(\'HF_TOKEN\')}""},\n json={""text"": ""hello""},\n timeout=60)\nprint(r.status_code, r.text)\nIf you want to implement more complex access control.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-26T09:10:43.033Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 22.0, 'yours': False, 'topic_id': 167772, 'topic_slug': 'private-space-authentication-for-external-api-calls', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/hub/en/spaces-embed', 'internal': False, 'reflection': False, 'title': 'Embed your Space in another website', 'clicks': 2}, {'url': 'https://huggingface.co/spaces/zero-gpu-explorers/README/discussions/88#68a736ebb21506a456c47c81', 'internal': False, 'reflection': False, 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/private-space-authentication-for-external-api-calls/167772/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 240278, 'name': 'Mohamed Nasr', 'username': 'nasr7322', 'avatar_template': '/user_avatar/discuss.huggingface.co/nasr7322/{size}/53080_2.png', 'created_at': '2025-08-26T09:11:44.798Z', 'cooked': 'yup it worked, thank youu!
\nmy problem was with the token
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-26T21:12:23.222Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 0.8, 'yours': False, 'topic_id': 167772, 'topic_slug': 'private-space-authentication-for-external-api-calls', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/private-space-authentication-for-external-api-calls/167772/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone!
+I’m using a Docker Space to deploy my FastAPI application that uses multiple models, but I’ve set it to private since my project contains sensitive code. My problem is that I can’t send requests to the endpoints from anywhere outside my browser and get a 404.
Is it possible to send a token with the request to authenticate myself? If so, how should I include it in my request to make it work properly?
Thank you all in advance!
If the space is functioning properly, you should be able to access it like following.
+You can figure out the actual space URL yourself, also you can also find it using the GUI.
curl -X POST https://OWNER-SPACENAME.hf.space/api/predict \
+ -H ""Authorization: Bearer $HF_TOKEN"" \
+ -H ""Content-Type: application/json"" \
+ -d '{""text"":""hello""}'
+or
+import os, requests
+url = ""https://OWNER-SPACENAME.hf.space/api/predict""
+r = requests.post(url,
+ headers={""Authorization"": f""Bearer {os.getenv('HF_TOKEN')}""},
+ json={""text"": ""hello""},
+ timeout=60)
+print(r.status_code, r.text)
+If you want to implement more complex access control.
" +Vet/vetgpt-2-7b n8n connection,https://discuss.huggingface.co/t/vet-vetgpt-2-7b-n8n-connection/167187,167187,5,2025-08-18 16:40:15.956000+00:00,"[{'id': 239110, 'name': 'Cristiane Sousa', 'username': 'ketask', 'avatar_template': '/user_avatar/discuss.huggingface.co/ketask/{size}/52727_2.png', 'created_at': '2025-08-18T16:40:16.017Z', 'cooked': 'Hi! I’m trying to connect HF model at N8N, but I receive error: “NodeOperationError: An error occurred while fetching the blob”. Is it due to I’m not using HF Pro plan?
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-18T16:40:16.017Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 3, 'readers_count': 2, 'score': 75.6, 'yours': False, 'topic_id': 167187, 'topic_slug': 'vet-vetgpt-2-7b-n8n-connection', 'display_username': 'Cristiane Sousa', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102003, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/vet-vetgpt-2-7b-n8n-connection/167187/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239200, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-19T04:36:31.730Z', 'cooked': 'That model location may be incorrect. Also, that model is not currently deployed, so it should not be available via the API.
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-19T04:36:31.730Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 167187, 'topic_slug': 'vet-vetgpt-2-7b-n8n-connection', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/models?inference_provider=all&sort=trending&search=vetgpt', 'internal': False, 'reflection': False, 'title': 'Models - Hugging Face', 'clicks': 1}, {'url': 'https://huggingface.co/ArcanaBT/vetgpt-2-7b', 'internal': False, 'reflection': False, 'title': 'ArcanaBT/vetgpt-2-7b · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/vet-vetgpt-2-7b-n8n-connection/167187/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 240301, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-26T13:15:40.680Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-26T13:15:40.680Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167187, 'topic_slug': 'vet-vetgpt-2-7b-n8n-connection', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/vet-vetgpt-2-7b-n8n-connection/167187/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi! I’m trying to connect HF model at N8N, but I receive error: “NodeOperationError: An error occurred while fetching the blob”. Is it due to I’m not using HF Pro plan?
+","That model location may be incorrect. Also, that model is not currently deployed, so it should not be available via the API.
" +Chat Templates for BlenderBot,https://discuss.huggingface.co/t/chat-templates-for-blenderbot/58184,58184,9,2023-10-11 14:56:57.572000+00:00,"[{'id': 93934, 'name': 'Rich Bergmann', 'username': 'bogolese', 'avatar_template': '/user_avatar/discuss.huggingface.co/bogolese/{size}/53040_2.png', 'created_at': '2023-10-11T14:56:57.642Z', 'cooked': 'I have installed transformers==4.34.0, tokenizers=0.14.1, and huggingface_hub=0.18.0 on Ubuntu 20 and I am trying to run the bog standard sample chat templates code from Templates for Chat Models under PyCharm. The error I consistently get is:
\nTraceback (most recent call last):
\nFile “/usr/local/lib/python3.8/dist-packages/IPython/core/interactiveshell.py”, line 3433, in run_code
\nexec(code_obj, self.user_global_ns, self.user_ns)
\nFile “”, line 10, in
\ntokenizer.apply_chat_template(chat, tokenize=False)
\nAttributeError: ‘BlenderbotTokenizerFast’ object has no attribute ‘apply_chat_template’
I need clues!
I generally solve this type of problem by asking chat-gpt. Just past your full code there and then add to the prompt the complete error (specifying the line) and ask for the correct code.
\nDon’t be afraid to ask if you have any problems.
Thanks, but this is not a syntax issue. It is an object model issue. Clearly there is an install dependency problem.
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2023-10-11T18:50:38.720Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 31, 'readers_count': 30, 'score': 36.2, 'yours': False, 'topic_id': 58184, 'topic_slug': 'chat-templates-for-blenderbot', 'display_username': 'Rich Bergmann', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 30826, 'username': 'Elciccio', 'name': 'Michele', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/7bcc69/{size}.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6790, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/chat-templates-for-blenderbot/58184/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 141327, 'name': 'Tarush Agarwal', 'username': 'hitarush', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/h/df788c/{size}.png', 'created_at': '2024-07-03T00:05:37.350Z', 'cooked': 'Hi, @bogolese, Did you manage to fix this dependancy issue?
', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2024-07-03T00:05:37.350Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 17, 'readers_count': 16, 'score': 63.4, 'yours': False, 'topic_id': 58184, 'topic_slug': 'chat-templates-for-blenderbot', 'display_username': 'Tarush Agarwal', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6790, 'username': 'bogolese', 'name': 'Rich Bergmann', 'avatar_template': '/user_avatar/discuss.huggingface.co/bogolese/{size}/53040_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 56360, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/chat-templates-for-blenderbot/58184/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 153032, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2024-09-02T08:38:28.374Z', 'cooked': 'Hi,
\nBlenderbot does not have a chat template set (there’s no “chat_template” attribute in the tokenizer_config.json). We’re going to update the docs to mention another model. cc @Rocketknight1
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2024-09-02T08:38:28.374Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 13, 'readers_count': 12, 'score': 42.6, 'yours': False, 'topic_id': 58184, 'topic_slug': 'chat-templates-for-blenderbot', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/chat-templates-for-blenderbot/58184/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 153034, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2024-09-02T08:54:55.948Z', 'cooked': 'Opened an issue here: ValueError: Cannot use apply_chat_template() because tokenizer.chat_template is not set · Issue #33246 · huggingface/transformers · GitHub
', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2024-09-02T08:54:55.948Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 13, 'readers_count': 12, 'score': 22.6, 'yours': False, 'topic_id': 58184, 'topic_slug': 'chat-templates-for-blenderbot', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/issues/33246', 'internal': False, 'reflection': False, 'title': 'ValueError: Cannot use apply_chat_template() because tokenizer.chat_template is not set · Issue #33246 · huggingface/transformers · GitHub', 'clicks': 54}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/chat-templates-for-blenderbot/58184/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 240226, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-25T16:11:42.043Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-08-25T16:11:42.043Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 58184, 'topic_slug': 'chat-templates-for-blenderbot', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/chat-templates-for-blenderbot/58184/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have installed transformers==4.34.0, tokenizers=0.14.1, and huggingface_hub=0.18.0 on Ubuntu 20 and I am trying to run the bog standard sample chat templates code from Templates for Chat Models under PyCharm. The error I consistently get is:
+Traceback (most recent call last):
+File “/usr/local/lib/python3.8/dist-packages/IPython/core/interactiveshell.py”, line 3433, in run_code
+exec(code_obj, self.user_global_ns, self.user_ns)
+File “”, line 10, in
+tokenizer.apply_chat_template(chat, tokenize=False)
+AttributeError: ‘BlenderbotTokenizerFast’ object has no attribute ‘apply_chat_template’
I need clues!
Opened an issue here: ValueError: Cannot use apply_chat_template() because tokenizer.chat_template is not set · Issue #33246 · huggingface/transformers · GitHub
" +HTTP Error 429 while running MMLU,https://discuss.huggingface.co/t/http-error-429-while-running-mmlu/167647,167647,5,2025-08-22 22:33:23.322000+00:00,"[{'id': 239977, 'name': 'Jay', 'username': 'jaydeepb', 'avatar_template': '/user_avatar/discuss.huggingface.co/jaydeepb/{size}/14906_2.png', 'created_at': '2025-08-22T22:33:23.379Z', 'cooked': 'Hi there. I’m trying to use the MMLU benchmark available at cais/mmlu · Datasets at Hugging Face . I have been trying to use it but running into HTTP Error 429 thrown while requesting HEAD ``https://huggingface.co/datasets/cais/mmlu/resolve/main/README.md. What could be the reason?
When error 429 occurs, it may be caused by IPv6, an outdated implementation of the old datasets library, or other factors.
\nIf it is truly an intentional rate limit, I believe only Hugging Face can resolve it…
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-23T00:11:12.478Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 26.6, 'yours': False, 'topic_id': 167647, 'topic_slug': 'http-error-429-while-running-mmlu', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/how-does-the-hub-handles-http-error-429/147346/3', 'internal': True, 'reflection': False, 'title': 'How does the hub handles http error 429?', 'clicks': 3}, {'url': 'https://github.com/huggingface/datasets/issues/7506', 'internal': False, 'reflection': False, 'title': 'HfHubHTTPError: 429 Client Error: Too Many Requests for URL when trying to access Fineweb-10BT on 4A100 GPUs using SLURM · Issue #7506 · huggingface/datasets · GitHub', 'clicks': 3}, {'url': 'https://github.com/huggingface/datasets/issues/7344#issuecomment-2582422510', 'internal': False, 'reflection': False, 'title': 'HfHubHTTPError: 429 Client Error: Too Many Requests for URL when trying to access SlimPajama-627B or c4 on TPUs · Issue #7344 · huggingface/datasets · GitHub', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/http-error-429-while-running-mmlu/167647/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 239987, 'name': 'Jay', 'username': 'jaydeepb', 'avatar_template': '/user_avatar/discuss.huggingface.co/jaydeepb/{size}/14906_2.png', 'created_at': '2025-08-23T03:55:14.848Z', 'cooked': '\n@John6666 thank you so much! using huggingface-cli login with my access token fixed this.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-23T15:55:23.410Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 167647, 'topic_slug': 'http-error-429-while-running-mmlu', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/http-error-429-while-running-mmlu/167647/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi there. I’m trying to use the MMLU benchmark available at cais/mmlu · Datasets at Hugging Face . I have been trying to use it but running into HTTP Error 429 thrown while requesting HEAD ``https://huggingface.co/datasets/cais/mmlu/resolve/main/README.md. What could be the reason?
When error 429 occurs, it may be caused by IPv6, an outdated implementation of the old datasets library, or other factors.
+If it is truly an intentional rate limit, I believe only Hugging Face can resolve it…
" +Is prometheus-eval not available on HuggingFace Spaces?,https://discuss.huggingface.co/t/is-prometheus-eval-not-available-on-huggingface-spaces/167309,167309,5,2025-08-19 18:24:25.866000+00:00,"[{'id': 239319, 'name': 'Hugo Torres', 'username': 'HugoFTorres', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugoftorres/{size}/52535_2.png', 'created_at': '2025-08-19T18:24:25.958Z', 'cooked': 'I am trying to use this library to evaluate my model, but whenever I add it to the requirements ile, I get a Build Error with the message:
\nERROR: Could not find a version that satisfies the requirement prometheus-eval (from versions: none) ERROR: No matching distribution found for prometheus-eval
\nIs there any step that I am missing here?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-19T18:24:25.958Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 167309, 'topic_slug': 'is-prometheus-eval-not-available-on-huggingface-spaces', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/is-prometheus-eval-not-available-on-huggingface-spaces/167309/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239374, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-20T01:59:38.030Z', 'cooked': 'It seems that Python version 3.10 to 3.12 is required for installation of prometheus-eval.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-23T14:49:27.194Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 167309, 'topic_slug': 'is-prometheus-eval-not-available-on-huggingface-spaces', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/is-prometheus-eval-not-available-on-huggingface-spaces/167309/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to use this library to evaluate my model, but whenever I add it to the requirements ile, I get a Build Error with the message:
+ERROR: Could not find a version that satisfies the requirement prometheus-eval (from versions: none) ERROR: No matching distribution found for prometheus-eval
+Is there any step that I am missing here?
","It seems that Python version 3.10 to 3.12 is required for installation of prometheus-eval.
Hello! I am fairly new to HuggingFace Spaces and I am trying to run an application, but keep getting the error [Errno 13] Permission denied: ‘/.streamlit’ . I have searched other topics and currently, even after setting HOME to /tmp/ or /data/. I have also added ENV PYTHONUNBUFFERED=1 \\ PORT=8000 \\ HF_HOME=/home/user/huggingface to the dockerfile, following another similar topic I have found but for some reason it doesn’t seem to run, or at least does not appear in the logs and I keep getting the same error on the container. Any idea on how to solve this?
', 'post_number': 1, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-08-13T09:54:30.243Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 343, 'reads': 8, 'readers_count': 7, 'score': 1571.4, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/permissionerror-errno-13-permission-denied-streamlit/166854/2', 'internal': True, 'reflection': True, 'title': ""PermissionError: [Errno 13] Permission denied: '/.streamlit'"", 'clicks': 5}, {'url': 'https://discuss.huggingface.co/t/space-stuck-on-starting-no-visible-logs-db-download-streamlit-app/166765/2', 'internal': True, 'reflection': True, 'title': 'Space stuck on “Starting” — no visible logs, DB download & Streamlit app', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 238285, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-13T10:13:48.706Z', 'cooked': 'There are some restrictions on directory access, so it is safer to refer to the official Docker sample. Also, the port to be used is written in README.md.
The final version looks like this.
', 'post_number': 2, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-08-13T10:13:48.706Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 6.4, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/John6666/streamlittest1', 'internal': False, 'reflection': False, 'title': 'Streamlittest1 - a Hugging Face Space by John6666', 'clicks': 24}, {'url': 'https://huggingface.co/docs/hub/en/spaces-sdks-docker-first-demo', 'internal': False, 'reflection': False, 'title': 'Your First Docker Space: Text Generation with T5', 'clicks': 22}, {'url': 'https://huggingface.co/docs/hub/en/spaces-config-reference', 'internal': False, 'reflection': False, 'title': 'Spaces Configuration Reference', 'clicks': 12}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238294, 'name': 'Hugo Torres', 'username': 'HugoFTorres', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugoftorres/{size}/52535_2.png', 'created_at': '2025-08-13T10:28:50.072Z', 'cooked': 'I have checked and it seems like we have the same configuration. However, the error persists and I still don’t understand why Would it help to provide the full log?
Hmm… My Dockerfile is just:
FROM python:3.9-slim\n\nWORKDIR /app\n\nRUN apt-get update && apt-get install -y \\\n build-essential \\\n curl \\\n git \\\n && rm -rf /var/lib/apt/lists/*\n\nCOPY requirements.txt ./\nCOPY src/ ./src/\n\nRUN pip3 install -r requirements.txt\n\nEXPOSE 8501\n\nHEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health\n\nENTRYPOINT [""streamlit"", ""run"", ""src/streamlit_app.py"", ""--server.port=8501"", ""--server.address=0.0.0.0""]\nAnd README.md:
---\ntitle: Streamlittest1\nemoji: 🚀\ncolorFrom: red\ncolorTo: red\nsdk: docker\napp_port: 8501\ntags:\n- streamlit\npinned: false\nshort_description: Streamlit template space\n---\nStrange, exact same as me. Meanwhile I figured out that my file_uploader was not working and figured I needed to create a .streamlit folder with a config.toml file inside it. Placed this folder at the root of the project and wondered if it couldn’t find it because it didn’t exist. However, after creating it, it still raises the same error. The app runs, but I believe this is messing with its correct functioning. Should this folder be in a different place? Are there any other configurations required?
', 'post_number': 5, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-08-13T11:21:25.582Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 13, 'reads': 8, 'readers_count': 7, 'score': 66.4, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238320, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-13T11:28:26.869Z', 'cooked': 'The root directory of the virtual machine that is executed is different from the root directory of the repository, so it would be better to modify the Dockerfile rather than the repository file structure.
For example, when specifying directories, it is better to write useradd first.
Ok, I kind of see the point of this, but can you help me understand how does this blends with the default dockerfile? Because it already contains commands such as WORKDIR. Should they be changed or is this something that should compliment what it already there?
', 'post_number': 7, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-08-13T11:46:36.715Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 8, 'readers_count': 7, 'score': 71.4, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238324, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-13T11:51:26.566Z', 'cooked': '\n\nShould they be changed or is this something that should compliment what it already there?
\n
Yeah. It seems to work fine that way.
', 'post_number': 8, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-08-13T11:51:26.566Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 8, 'readers_count': 7, 'score': 21.4, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/hub/en/spaces-sdks-docker-first-demo#create-the-dockerfile', 'internal': False, 'reflection': False, 'title': 'Your First Docker Space: Text Generation with T5', 'clicks': 48}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 238334, 'name': 'Hugo Torres', 'username': 'HugoFTorres', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugoftorres/{size}/52535_2.png', 'created_at': '2025-08-13T12:47:14.690Z', 'cooked': 'Added the user part and it seems to be working! I get a completely different error, but it is something for another topic. Thank you for your help!
', 'post_number': 9, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-08-13T12:47:14.690Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 8, 'readers_count': 7, 'score': 26.4, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/9', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 240039, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-23T14:49:27.193Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 10, 'post_type': 3, 'posts_count': 10, 'updated_at': '2025-08-23T14:49:27.193Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 5.8, 'yours': False, 'topic_id': 166664, 'topic_slug': 'i-keep-getting-errno-13-permission-denied-streamlit', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/i-keep-getting-errno-13-permission-denied-streamlit/166664/10', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello! I am fairly new to HuggingFace Spaces and I am trying to run an application, but keep getting the error [Errno 13] Permission denied: ‘/.streamlit’ . I have searched other topics and currently, even after setting HOME to /tmp/ or /data/. I have also added ENV PYTHONUNBUFFERED=1 \ PORT=8000 \ HF_HOME=/home/user/huggingface to the dockerfile, following another similar topic I have found but for some reason it doesn’t seem to run, or at least does not appear in the logs and I keep getting the same error on the container. Any idea on how to solve this?
","++Should they be changed or is this something that should compliment what it already there?
+
Yeah. It seems to work fine that way.
" +Space currently stuck on building,https://discuss.huggingface.co/t/space-currently-stuck-on-building/167637,167637,5,2025-08-22 15:36:30.234000+00:00,"[{'id': 239953, 'name': 'Hugo Torres', 'username': 'HugoFTorres', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugoftorres/{size}/52535_2.png', 'created_at': '2025-08-22T15:36:30.317Z', 'cooked': 'Hello! My space is currently stuck at building after a couple of changes. It doesn’t even produc any logs. I have seen older topics in which the same was pointed out but it was a HuggingFace issue. Is there any way I can validate if it is s Spaces issue or an issue of my specific space?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-22T15:36:30.317Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 6, 'readers_count': 5, 'score': 41.2, 'yours': False, 'topic_id': 167637, 'topic_slug': 'space-currently-stuck-on-building', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-currently-stuck-on-building/167637/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239979, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-22T23:42:57.257Z', 'cooked': 'There is no official way to confirm whether this issue or not…
\nAs a workaround, try creating a new space and uploading the same source code to see if it works.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-23T14:48:27.674Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 167637, 'topic_slug': 'space-currently-stuck-on-building', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/space-currently-stuck-on-building/167637/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]",Hello! My space is currently stuck at building after a couple of changes. It doesn’t even produc any logs. I have seen older topics in which the same was pointed out but it was a HuggingFace issue. Is there any way I can validate if it is s Spaces issue or an issue of my specific space?
,"There is no official way to confirm whether this issue or not…
+As a workaround, try creating a new space and uploading the same source code to see if it works.
Hello huggingface community. I am wondering if I did understand the pipeline text-classification correctly. Is it the case, that the model I choose defines the task I can do with it and the output I will get? I was a bit confused, because I used pipeline(“sentiment-analysis”) but did not find “sentiment-analysis” as a model or option setting. And VSCode autocomplete also did not suggest it, but it still works. So I came to the conclusion I laid out before. Is this correct or am I wrong. Thanks and may you all have a good time.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-22T19:06:44.198Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 7, 'readers_count': 6, 'score': 71.4, 'yours': False, 'topic_id': 167640, 'topic_slug': 'text-classification-pipeline-newbie-question', 'display_username': 'Markus Eicher', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 29747, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-classification-pipeline-newbie-question/167640/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239972, 'name': 'Daniel Kleine', 'username': 'dkleine', 'avatar_template': '/user_avatar/discuss.huggingface.co/dkleine/{size}/33964_2.png', 'created_at': '2025-08-22T19:51:01.268Z', 'cooked': 'Hi Markus,
\n“sentiment-analysis” is the task specifying what you want a large language model to perform on the text. Sentiment analysis practically changes the model’s head to a classifier, which you can see here:
This pipeline is pre-configured, the settings can be found below in the same file defined here:
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-22T19:51:27.289Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 36.4, 'yours': False, 'topic_id': 167640, 'topic_slug': 'text-classification-pipeline-newbie-question', 'display_username': 'Daniel Kleine', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/blob/7d88f57fc6892b9b3d0092c53e27ae033f1bebc8/src/transformers/pipelines/__init__.py#L193-L205', 'internal': False, 'reflection': False, 'title': 'transformers/src/transformers/pipelines/__init__.py at 7d88f57fc6892b9b3d0092c53e27ae033f1bebc8 · huggingface/transformers · GitHub', 'clicks': 1}, {'url': 'https://github.com/huggingface/transformers/blob/7d88f57fc6892b9b3d0092c53e27ae033f1bebc8/src/transformers/pipelines/__init__.py#L154-L159', 'internal': False, 'reflection': False, 'title': 'transformers/src/transformers/pipelines/__init__.py at 7d88f57fc6892b9b3d0092c53e27ae033f1bebc8 · huggingface/transformers · GitHub', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/default-models-for-pipeline-tasks/2559/6', 'internal': True, 'reflection': True, 'title': 'Default models for pipeline tasks', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 69473, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-classification-pipeline-newbie-question/167640/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 239973, 'name': 'Markus Eicher', 'username': 'MarkusEicher', 'avatar_template': '/user_avatar/discuss.huggingface.co/markuseicher/{size}/52883_2.png', 'created_at': '2025-08-22T20:11:08.187Z', 'cooked': 'Thank you. So it is generally an alias for text-classification. I was confused because it did not show up as a separate pipeline in chapter 1 of the LLM course on huggingface. But now I understand why. Appreciate your support and the quick answer.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-22T20:11:08.187Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 6, 'readers_count': 5, 'score': 56.2, 'yours': False, 'topic_id': 167640, 'topic_slug': 'text-classification-pipeline-newbie-question', 'display_username': 'Markus Eicher', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 69473, 'username': 'dkleine', 'name': 'Daniel Kleine', 'avatar_template': '/user_avatar/discuss.huggingface.co/dkleine/{size}/33964_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 29747, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-classification-pipeline-newbie-question/167640/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 239974, 'name': 'Daniel Kleine', 'username': 'dkleine', 'avatar_template': '/user_avatar/discuss.huggingface.co/dkleine/{size}/33964_2.png', 'created_at': '2025-08-22T20:23:18.891Z', 'cooked': 'That’s right – “sentiment-analysis” practically does sequence classification (there are also other types of classification tasks possible, for example token classification, just fyi) under the hood in the linear output layer of the LLM. Please also see the docstring for the TextClassificationPipeline here:
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-08-23T08:23:30.049Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 167640, 'topic_slug': 'text-classification-pipeline-newbie-question', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/text-classification-pipeline-newbie-question/167640/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello huggingface community. I am wondering if I did understand the pipeline text-classification correctly. Is it the case, that the model I choose defines the task I can do with it and the output I will get? I was a bit confused, because I used pipeline(“sentiment-analysis”) but did not find “sentiment-analysis” as a model or option setting. And VSCode autocomplete also did not suggest it, but it still works. So I came to the conclusion I laid out before. Is this correct or am I wrong. Thanks and may you all have a good time.
","Hi Markus,
+“sentiment-analysis” is the task specifying what you want a large language model to perform on the text. Sentiment analysis practically changes the model’s head to a classifier, which you can see here:
This pipeline is pre-configured, the settings can be found below in the same file defined here:
+ +" +ImportError: cannot import name ‘ModelFilter’ from ‘huggingface_hub’,https://discuss.huggingface.co/t/importerror-cannot-import-name-modelfilter-from-huggingface-hub/167632,167632,5,2025-08-22 13:18:09.224000+00:00,"[{'id': 239912, 'name': 'Alex', 'username': 'SuperBowser', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9f8e36/{size}.png', 'created_at': '2025-08-22T13:18:09.284Z', 'cooked': 'I am running this line in Kaggle notebook:
\nfrom huggingface_hub import ModelFilter\nand getting back error:
\n---------------------------------------------------------------------------\nImportError Traceback (most recent call last)\n/tmp/ipykernel_36/1451250264.py in <cell line: 0>()\n----> 1 from huggingface_hub import ModelFilter\n\nImportError: cannot import name \'ModelFilter\' from \'huggingface_hub\' (/usr/local/lib/python3.11/dist-packages/huggingface_hub/__init__.py)\nMy huggingface_hub._version_ is ‘0.33.1’
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-22T13:18:09.284Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 108, 'reads': 6, 'readers_count': 5, 'score': 481.2, 'yours': False, 'topic_id': 167632, 'topic_slug': 'importerror-cannot-import-name-modelfilter-from-huggingface-hub', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/importerror-cannot-import-name-modelfilter-from-huggingface-hub/167632/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239950, 'name': 'Daniel Kleine', 'username': 'dkleine', 'avatar_template': '/user_avatar/discuss.huggingface.co/dkleine/{size}/33964_2.png', 'created_at': '2025-08-22T15:21:25.382Z', 'cooked': 'ModelFilter is deprecated, please see here: ImportError: cannot import name \'ModelFilter\' from \'huggingface_hub\' · Issue #2478 · huggingface/huggingface_hub · GitHub
Thank you so much for your answer. Do you what values I can use in filter field. I am looking for complete list. So far I know only a few values such text-classification
Minor update. Here is my search:
\nfrom huggingface_hub import HfApi
\napi = HfApi()
\nmodels = api.list_models(task=“text-classification”,
\nsort=‘downloads’, gated = False, limit = 100)
\nmodels = list(models)
\nprint(len(models))
\nprint(models[1].modelId)
It returns cross-encoder/ms-marco-MiniLM-L6-v2, which is “Text Ranking” and it is different from what I asked “Text Classification” as per tasks page.
\nI got the same result when using “filter” field.
\n\nIt returns
\ncross-encoder/ms-marco-MiniLM-L6-v2, which is “Text Ranking” and it is different from what I asked “Text Classification” as per tasks page.
\nI got the same result when using “filter” field.
This is probably because this model is tagged as both as “Text Ranking” as well as “Text Classification”, see tags above:
\n\n', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-22T19:08:35.289Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 4, 'readers_count': 3, 'score': 55.8, 'yours': False, 'topic_id': 167632, 'topic_slug': 'importerror-cannot-import-name-modelfilter-from-huggingface-hub', 'display_username': 'Daniel Kleine', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/cross-encoder/ms-marco-MiniLM-L6-v2', 'internal': False, 'reflection': False, 'title': 'cross-encoder/ms-marco-MiniLM-L6-v2 · Hugging Face', 'clicks': 1}, {'url': 'https://huggingface.co/tasks', 'internal': False, 'reflection': False, 'title': 'Tasks - Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 69473, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/importerror-cannot-import-name-modelfilter-from-huggingface-hub/167632/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 239997, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-23T07:07:27.219Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-08-23T07:07:27.219Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 167632, 'topic_slug': 'importerror-cannot-import-name-modelfilter-from-huggingface-hub', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/importerror-cannot-import-name-modelfilter-from-huggingface-hub/167632/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am running this line in Kaggle notebook:
+from huggingface_hub import ModelFilter
+and getting back error:
+---------------------------------------------------------------------------
+ImportError Traceback (most recent call last)
+/tmp/ipykernel_36/1451250264.py in <cell line: 0>()
+----> 1 from huggingface_hub import ModelFilter
+
+ImportError: cannot import name 'ModelFilter' from 'huggingface_hub' (/usr/local/lib/python3.11/dist-packages/huggingface_hub/__init__.py)
+My huggingface_hub._version_ is ‘0.33.1’
","ModelFilter is deprecated, please see here: ImportError: cannot import name 'ModelFilter' from 'huggingface_hub' · Issue #2478 · huggingface/huggingface_hub · GitHub
I am able to download the Reddit-TIFU dataset,
\n\n\n\n
reddit_tifu = load_dataset(\'reddit_tifu\', \'long\', split=\'train\', trust_remote_code=True)
I have also used the dataset in the past and was able to access its dataset card (https://huggingface.co/reddit_tifu/datasets), but it now returns a 404 error. Is there a reason for this?
\nhttps://huggingface.co/reddit_tifu/datasets
', 'post_number': 1, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-20T15:01:21.327Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 5, 'readers_count': 4, 'score': 51.0, 'yours': False, 'topic_id': 167436, 'topic_slug': 'missing-dataset-card-reddit-tifu-dataset', 'display_username': 'Anna Kougioumtzidou', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/reddit_tifu/datasets', 'internal': False, 'reflection': False, 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 10170, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/missing-dataset-card-reddit-tifu-dataset/167436/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239658, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-21T01:52:08.018Z', 'cooked': 'It appears that the user does not exist at this time.
', 'post_number': 2, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-21T01:52:08.018Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 167436, 'topic_slug': 'missing-dataset-card-reddit-tifu-dataset', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/reddit_tifu', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/missing-dataset-card-reddit-tifu-dataset/167436/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 239757, 'name': 'Anna Kougioumtzidou', 'username': 'Anna-Kay', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/97f17d/{size}.png', 'created_at': '2025-08-21T10:52:13.865Z', 'cooked': 'Thanks for the quick response!
\nDoes this mean that the dataset itself may go missing in the future? Should I file an issue?
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-21T10:52:13.865Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 167436, 'topic_slug': 'missing-dataset-card-reddit-tifu-dataset', 'display_username': 'Anna Kougioumtzidou', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 10170, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/missing-dataset-card-reddit-tifu-dataset/167436/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 239763, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-21T12:21:09.083Z', 'cooked': 'Oh, sorry, I just found it now.
load_dataset(\'reddit_tifu\')
\nIn this case, the user name is automatically completed. Therefore, you need to search to find the actual link.
Thanks a lot for this!
', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-22T13:21:28.325Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 167436, 'topic_slug': 'missing-dataset-card-reddit-tifu-dataset', 'display_username': 'Anna Kougioumtzidou', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 10170, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/missing-dataset-card-reddit-tifu-dataset/167436/6', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 239982, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-23T01:21:29.099Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-08-23T01:21:29.099Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 167436, 'topic_slug': 'missing-dataset-card-reddit-tifu-dataset', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/missing-dataset-card-reddit-tifu-dataset/167436/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am able to download the Reddit-TIFU dataset,
++++
reddit_tifu = load_dataset('reddit_tifu', 'long', split='train', trust_remote_code=True)
I have also used the dataset in the past and was able to access its dataset card (https://huggingface.co/reddit_tifu/datasets), but it now returns a 404 error. Is there a reason for this?
+https://huggingface.co/reddit_tifu/datasets
","Oh, sorry, I just found it now.
Hi, I’m trying to run the second setup line for the RL Course, Unit 1:
\npip install -r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt\nHowever, I get the following error:
\n...\nCollecting pygame==2.1.3 (from gymnasium[box2d]->-r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt (line 3))\n Using cached pygame-2.1.3.tar.gz (12.8 MB)\n error: subprocess-exited-with-error\n \n × python setup.py egg_info did not run successfully.\n │ exit code: 1\n ╰─> See above for output.\n \n note: This error originates from a subprocess, and is likely not a problem with pip.\n Preparing metadata (setup.py) ... error\nerror: metadata-generation-failed\n\n× Encountered error while generating package metadata.\n╰─> See above for output.\n\nnote: This is an issue with the package mentioned above, not pip.\nhint: See above for details.\nI’ve tried solutions from other question threads and can’t seem to resolve this.
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-20T14:05:25.487Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 213, 'reads': 13, 'readers_count': 12, 'score': 982.6, 'yours': False, 'topic_id': 167429, 'topic_slug': 'rl-course-unit-1-python-setup-py-egg-info-did-not-run-successfully', 'display_username': 'Pearl Yu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102149, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/rl-course-unit-1-python-setup-py-egg-info-did-not-run-successfully/167429/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239491, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-20T14:41:02.295Z', 'cooked': 'stable-baselines3==2.0.0a5\nswig\ngymnasium[box2d]\nhuggingface_sb3\nIt seems that there is a problem with box2d with the gymnasium library to be installed there.
Thanks for your response! It definitely led me in the right direction. Essentially I replaced the line
\n!pip install -r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt\nwith the following lines
\n!pip install stable-baselines3==2.0.0a5\n!pip install swig\n!pip install gymnasium\n!pip install box2d-py\n!pip install huggingface_sb3\nwhich does not err and appears to install the same necessary components.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-20T17:19:03.526Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 13, 'readers_count': 12, 'score': 122.6, 'yours': False, 'topic_id': 167429, 'topic_slug': 'rl-course-unit-1-python-setup-py-egg-info-did-not-run-successfully', 'display_username': 'Pearl Yu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102149, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/rl-course-unit-1-python-setup-py-egg-info-did-not-run-successfully/167429/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 3}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 239683, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-21T05:19:42.039Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-21T05:19:42.039Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 12, 'readers_count': 11, 'score': 32.4, 'yours': False, 'topic_id': 167429, 'topic_slug': 'rl-course-unit-1-python-setup-py-egg-info-did-not-run-successfully', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/rl-course-unit-1-python-setup-py-egg-info-did-not-run-successfully/167429/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi, I’m trying to run the second setup line for the RL Course, Unit 1:
+pip install -r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt
+However, I get the following error:
+...
+Collecting pygame==2.1.3 (from gymnasium[box2d]->-r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt (line 3))
+ Using cached pygame-2.1.3.tar.gz (12.8 MB)
+ error: subprocess-exited-with-error
+
+ × python setup.py egg_info did not run successfully.
+ │ exit code: 1
+ ╰─> See above for output.
+
+ note: This error originates from a subprocess, and is likely not a problem with pip.
+ Preparing metadata (setup.py) ... error
+error: metadata-generation-failed
+
+× Encountered error while generating package metadata.
+╰─> See above for output.
+
+note: This is an issue with the package mentioned above, not pip.
+hint: See above for details.
+I’ve tried solutions from other question threads and can’t seem to resolve this.
","Thanks for your response! It definitely led me in the right direction. Essentially I replaced the line
+!pip install -r https://raw.githubusercontent.com/huggingface/deep-rl-class/main/notebooks/unit1/requirements-unit1.txt
+with the following lines
+!pip install stable-baselines3==2.0.0a5
+!pip install swig
+!pip install gymnasium
+!pip install box2d-py
+!pip install huggingface_sb3
+which does not err and appears to install the same necessary components.
" +LORA - how to determine what module_to_save,https://discuss.huggingface.co/t/lora-how-to-determine-what-module-to-save/167206,167206,5,2025-08-18 19:38:10.239000+00:00,"[{'id': 239154, 'name': 'Alex', 'username': 'SuperBowser', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9f8e36/{size}.png', 'created_at': '2025-08-18T19:38:10.297Z', 'cooked': 'I am reading through LORA tutorial and one of the options in LoraConfig is modue_to_save. In the example its value is ‘decode-head’. I would like to use LORA with SequenceClassification model and I not sure what module I need to save.
\nAny thoughts?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-18T19:38:10.297Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 21, 'reads': 7, 'readers_count': 6, 'score': 86.4, 'yours': False, 'topic_id': 167206, 'topic_slug': 'lora-how-to-determine-what-module-to-save', 'display_username': 'Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/peft/main/en/task_guides/semantic_segmentation_lora', 'internal': False, 'reflection': False, 'title': 'Semantic segmentation using LoRA', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 102016, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/lora-how-to-determine-what-module-to-save/167206/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239206, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-19T05:35:21.233Z', 'cooked': 'If you specify task_type, PEFT will automatically set module_to_save to an appropriate value. If you want to manually search for the head module to save, it would look something like this.
from transformers import AutoModelForSequenceClassification, AutoConfig\nimport torch.nn as nn\n\nHEAD_CANDIDATES = (""classifier"", ""score"", ""logits_proj"", ""classification_head"")\n\ndef find_cls_head_name(model):\n present = [n for n, _ in model.named_modules() if n.split(""."")[-1] in HEAD_CANDIDATES]\n if present: return present[0], present\n num_labels = getattr(getattr(model, ""config"", object()), ""num_labels"", None)\n hits = []\n for parent_name, module in model.named_modules():\n for child_name, child in module.named_children():\n if isinstance(child, nn.Linear) and getattr(child, ""out_features"", None) == num_labels:\n hits.append(child_name if parent_name == """" else f""{parent_name}.{child_name}"")\n return (hits[0] if hits else None), hits\n\ndef print_head_name(model_name):\n cfg = AutoConfig.from_pretrained(model_name)\n model = AutoModelForSequenceClassification.from_pretrained(model_name, config=cfg)\n best, all_hits = find_cls_head_name(model)\n print(""Model name:"", model_name)\n print(""All candidate heads:"", all_hits)\n print(""Suggested modules_to_save:"", [best] if best else None)\n\nprint_head_name(""distilbert-base-uncased-finetuned-sst-2-english"")\n#Model name: distilbert-base-uncased-finetuned-sst-2-english\n#All candidate heads: [\'classifier\']\n#Suggested modules_to_save: [\'classifier\']\nprint_head_name(""HuggingFaceTB/SmolLM-135M"")\n#Model name: HuggingFaceTB/SmolLM-135M\n#All candidate heads: [\'score\']\n#Suggested modules_to_save: [\'score\']\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-20T19:27:47.311Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 167206, 'topic_slug': 'lora-how-to-determine-what-module-to-save', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/lora-how-to-determine-what-module-to-save/167206/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am reading through LORA tutorial and one of the options in LoraConfig is modue_to_save. In the example its value is ‘decode-head’. I would like to use LORA with SequenceClassification model and I not sure what module I need to save.
+Any thoughts?
","If you specify task_type, PEFT will automatically set module_to_save to an appropriate value. If you want to manually search for the head module to save, it would look something like this.
from transformers import AutoModelForSequenceClassification, AutoConfig
+import torch.nn as nn
+
+HEAD_CANDIDATES = (""classifier"", ""score"", ""logits_proj"", ""classification_head"")
+
+def find_cls_head_name(model):
+ present = [n for n, _ in model.named_modules() if n.split(""."")[-1] in HEAD_CANDIDATES]
+ if present: return present[0], present
+ num_labels = getattr(getattr(model, ""config"", object()), ""num_labels"", None)
+ hits = []
+ for parent_name, module in model.named_modules():
+ for child_name, child in module.named_children():
+ if isinstance(child, nn.Linear) and getattr(child, ""out_features"", None) == num_labels:
+ hits.append(child_name if parent_name == """" else f""{parent_name}.{child_name}"")
+ return (hits[0] if hits else None), hits
+
+def print_head_name(model_name):
+ cfg = AutoConfig.from_pretrained(model_name)
+ model = AutoModelForSequenceClassification.from_pretrained(model_name, config=cfg)
+ best, all_hits = find_cls_head_name(model)
+ print(""Model name:"", model_name)
+ print(""All candidate heads:"", all_hits)
+ print(""Suggested modules_to_save:"", [best] if best else None)
+
+print_head_name(""distilbert-base-uncased-finetuned-sst-2-english"")
+#Model name: distilbert-base-uncased-finetuned-sst-2-english
+#All candidate heads: ['classifier']
+#Suggested modules_to_save: ['classifier']
+print_head_name(""HuggingFaceTB/SmolLM-135M"")
+#Model name: HuggingFaceTB/SmolLM-135M
+#All candidate heads: ['score']
+#Suggested modules_to_save: ['score']
+ I just published the first part of a tutorial series on the Muon Optimizer.
Muon (Momentum Orthogonalized by Newton-Schulz) is quickly becoming the go-to optimizer for large-scale training. It’s already powering trillion-parameter frontier models like Kimi-2 (MuonClip) and was critical for the ATLAS paper, where first-order optimizers failed.
\nIn this series, I’m breaking Muon down step by step: intuition, pseudocode, PyTorch implementation, and practical guidance on when/where to use it.
\n\nAlso — I’d really like to contribute this as a guest article to the Hugging Face blog. I know the blog is managed by a group, but it looks like external contributors can’t directly join. If anyone here has advice or connections on how to submit contributions, I’d love to hear it
Muon deserves more attention in the open-source community, and I’d be excited to help bridge that gap.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-19T02:06:50.801Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 91, 'reads': 6, 'readers_count': 5, 'score': 456.2, 'yours': False, 'topic_id': 167227, 'topic_slug': 'first-instalment-the-muon-optimizer-tutorial-series', 'display_username': 'Jen Wei', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://medium.com/@jenwei0312/going-beyond-adamw-a-practical-guide-to-the-muon-optimizer-93d90e91dbd3', 'internal': False, 'reflection': False, 'title': 'Going Beyond AdamW: A Practical Guide to the Muon Optimizer | by Jennifer Wei | Aug, 2025 | Medium', 'clicks': 18}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 75338, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/first-instalment-the-muon-optimizer-tutorial-series/167227/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 239217, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-19T07:14:16.315Z', 'cooked': 'It seems that the standard procedure is to press the join button and wait for approval, or to post on GitHub. If you are in a hurry, it may be quicker to contact the staff via email or Discord. website@huggingface.co
\n
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-20T00:04:56.146Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 167227, 'topic_slug': 'first-instalment-the-muon-optimizer-tutorial-series', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/first-instalment-the-muon-optimizer-tutorial-series/167227/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]"," I just published the first part of a tutorial series on the Muon Optimizer.
Muon (Momentum Orthogonalized by Newton-Schulz) is quickly becoming the go-to optimizer for large-scale training. It’s already powering trillion-parameter frontier models like Kimi-2 (MuonClip) and was critical for the ATLAS paper, where first-order optimizers failed.
+In this series, I’m breaking Muon down step by step: intuition, pseudocode, PyTorch implementation, and practical guidance on when/where to use it.
+ +Also — I’d really like to contribute this as a guest article to the Hugging Face blog. I know the blog is managed by a group, but it looks like external contributors can’t directly join. If anyone here has advice or connections on how to submit contributions, I’d love to hear it
Muon deserves more attention in the open-source community, and I’d be excited to help bridge that gap.
","It seems that the standard procedure is to press the join button and wait for approval, or to post on GitHub. If you are in a hurry, it may be quicker to contact the staff via email or Discord. website@huggingface.co
+
i was trying to build a small AI agent that would query the DB and get the details of the customers, for which i tried many models that are available in the ollama model library, but every model keeps throwing an “invalid tool”, or keeps using the irrelevant tool or keeps hallucinating and giving back made up answers!!! is this an issue that is common when pulling and running LLM’s locally using OLLAMA, when i use the paid Gemini API from google cloud, it works so well (uses the correct tool’s, and returns the exact correct answer), i need help in understanding what is happening when i use a locally run LLM, and is there anyway to make the Local LLM work like the Gemini API??
\nThanks in advance
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-01T11:20:02.900Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 109, 'reads': 5, 'readers_count': 4, 'score': 536.0, 'yours': False, 'topic_id': 165277, 'topic_slug': 'tool-function-calling-abilities-of-llms-that-are-used-locally-pulled-through-ollama', 'display_username': 'Aravindha Sivabalan J', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100794, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/tool-function-calling-abilities-of-llms-that-are-used-locally-pulled-through-ollama/165277/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 235983, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-01T14:01:03.637Z', 'cooked': 'If you are using Ollama directly without any Agent framework, the models that support tool calling are limited, and there seems to be an issue that is not a bug.
\nAs a workaround, you could use Ollama through external Agent frameworks.
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-08-01T14:01:03.637Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 5, 'readers_count': 4, 'score': 46.0, 'yours': False, 'topic_id': 165277, 'topic_slug': 'tool-function-calling-abilities-of-llms-that-are-used-locally-pulled-through-ollama', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/how-to-run-agents-from-smolagents-locally/152874/3', 'internal': True, 'reflection': False, 'title': 'How to run agents from `smolagents` locally?', 'clicks': 12}, {'url': 'https://ollama.com/blog/tool-support', 'internal': False, 'reflection': False, 'title': 'Tool support · Ollama Blog', 'clicks': 9}, {'url': 'https://huggingface.co/posts/prithivMLmods/142876386338407', 'internal': False, 'reflection': False, 'title': '@prithivMLmods on Hugging Face: ""OpenAI, Google, Hugging Face, and Anthropic have released guides and courses…""', 'clicks': 7}, {'url': 'https://github.com/ollama/ollama/issues/11538', 'internal': False, 'reflection': False, 'title': 'Qwen3:14b not usingThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-19T09:27:01.360Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 165277, 'topic_slug': 'tool-function-calling-abilities-of-llms-that-are-used-locally-pulled-through-ollama', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/tool-function-calling-abilities-of-llms-that-are-used-locally-pulled-through-ollama/165277/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","i was trying to build a small AI agent that would query the DB and get the details of the customers, for which i tried many models that are available in the ollama model library, but every model keeps throwing an “invalid tool”, or keeps using the irrelevant tool or keeps hallucinating and giving back made up answers!!! is this an issue that is common when pulling and running LLM’s locally using OLLAMA, when i use the paid Gemini API from google cloud, it works so well (uses the correct tool’s, and returns the exact correct answer), i need help in understanding what is happening when i use a locally run LLM, and is there anyway to make the Local LLM work like the Gemini API??
+Thanks in advance
","If you are using Ollama directly without any Agent framework, the models that support tool calling are limited, and there seems to be an issue that is not a bug.
+As a workaround, you could use Ollama through external Agent frameworks.
" +QLoRA Fine-tuning is Too Slow on LLaMA-based Model Despite BitsAndBytes Optimization,https://discuss.huggingface.co/t/qlora-fine-tuning-is-too-slow-on-llama-based-model-despite-bitsandbytes-optimization/166964,166964,6,2025-08-16 10:05:35.466000+00:00,"[{'id': 238766, 'name': 'Aylin Naebzadeh', 'username': 'AylinNaebzadeh', 'avatar_template': '/user_avatar/discuss.huggingface.co/aylinnaebzadeh/{size}/52343_2.png', 'created_at': '2025-08-16T10:05:35.536Z', 'cooked': 'Hi everyone,
\nI’m fine-tuning a LLaMA-based model (universitytehran/PersianMind-v1.0) using QLoRA and BitsAndBytes in 4-bit precision. I am working with Kaggle GPU T4, and it takes about 75 hours to be fine-tuned using ParsMap dataset with 40,000 records for training related to converting informal to formal text.
\nHere is my code:
base_model_id = ""universitytehran/PersianMind-v1.0""\ncompute_dtype = torch.bfloat16 if torch.cuda.get_device_capability(0)[0] >= 8 else torch.float16\n\nprint(""Compute dtype:"", compute_dtype)\ndef safe_str(x):\n return """" if x is None or (isinstance(x, float) and np.isnan(x)) else str(x)\n\ndf = df_parsmap.copy()\ndf = df.dropna(subset=[""inFormalForm"",""formalForm""]) # keep only rows with both sides\n\ndef make_text(row):\n informal = safe_str(row[""inFormalForm""])\n formal = safe_str(row[""formalForm""])\n return f""<s><|startoftext|>[Informal]{informal}[Formal]{formal}<|endoftext|>""\n\ndf[""text""] = df.apply(make_text, axis=1)\nperm = np.random.permutation(len(df))\ncut = int(0.9*len(df))\ntrain_df = df.iloc[perm[:cut]].reset_index(drop=True)\nval_df = df.iloc[perm[cut:]].reset_index(drop=True)\n\nds = DatasetDict({\n ""train"": Dataset.from_pandas(train_df[[""text""]]),\n ""validation"": Dataset.from_pandas(val_df[[""text""]]),\n})\nlen(ds[""train""]), len(ds[""validation""])\n\ntokenizer = AutoTokenizer.from_pretrained(base_model_id, use_fast=True, trust_remote_code=True)\n\nspecials = {\n ""bos_token"": ""<s>"",\n ""eos_token"": ""</s>"",\n ""pad_token"": ""<pad>"",\n}\n\nfor k,v in specials.items():\n if getattr(tokenizer, k, None) != v:\n tokenizer.add_special_tokens({k: v})\n\nadded = tokenizer.add_tokens([""<|startoftext|>"", ""<|endoftext|>"", ""[Informal]"", ""[Formal]"", ""<sep>""], special_tokens=True)\nprint(""Added new tokens:"", added)\n\n\nif tokenizer.pad_token is None:\n tokenizer.pad_token = tokenizer.eos_token\nbnb_config = BitsAndBytesConfig(\n load_in_4bit=True,\n bnb_4bit_quant_type=""nf4"",\n bnb_4bit_use_double_quant=True,\n bnb_4bit_compute_dtype=compute_dtype,\n)\n\nmodel = AutoModelForCausalLM.from_pretrained(\n base_model_id,\n trust_remote_code=True,\n quantization_config=bnb_config,\n device_map=""auto"",\n)\n\nmodel.resize_token_embeddings(len(tokenizer))\n\nmodel = prepare_model_for_kbit_training(model)\nmodel.config.use_cache = False\nlora_config = LoraConfig(\n r=16, lora_alpha=32, lora_dropout=0.1, bias=""none"", task_type=""CAUSAL_LM"",\n target_modules=[""q_proj"",""k_proj"",""v_proj"",""o_proj"",""gate_proj"",""up_proj"",""down_proj""],\n)\nmodel = get_peft_model(model, lora_config)\n\nmodel.gradient_checkpointing_enable()\n\n# quick param report\ntrainable = sum(p.numel() for p in model.parameters() if p.requires_grad)\ntotal = sum(p.numel() for p in model.parameters())\nprint(f""Trainable: {trainable:,} / Total: {total:,} ({100*trainable/total:.2f}%)"")\nmax_length = 128\n\ndef tokenize_batch(batch):\n return tokenizer(\n batch[""text""],\n truncation=True,\n max_length=max_length,\n padding=""max_length"",\n )\n\ntokenized = ds.map(tokenize_batch, batched=True, remove_columns=ds[""train""].column_names)\ncollator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)
effective_bs = 16 \nper_device_train_bs = 2\nper_device_eval_bs = 2\ngrad_accum = max(1, effective_bs // per_device_train_bs)\nepochs = 3\n\nargs = TrainingArguments(\n output_dir=""./persianmind-formalizer-lora"",\n num_train_epochs=epochs,\n per_device_train_batch_size=per_device_train_bs,\n per_device_eval_batch_size=per_device_eval_bs,\n gradient_accumulation_steps=grad_accum,\n learning_rate=1e-5,\n warmup_ratio=0.03,\n lr_scheduler_type=""cosine"",\n weight_decay=0.0,\n logging_steps=50,\n\n eva_strategy=""steps"",\n eval_steps=2000, \n save_strategy=""epoch"", \n save_total_limit=2,\n load_best_model_at_end=True,\n\n bf16=(compute_dtype==torch.bfloat16),\n fp16=(compute_dtype==torch.float16),\n\n optim=""paged_adamw_8bit"", \n gradient_checkpointing=True,\n gradient_checkpointing_kwargs={""use_reentrant"": False},\n\n dataloader_num_workers=4,\n dataloader_pin_memory=True,\n dataloader_persistent_workers=True,\n\n group_by_length=True, \n tf32=True,\n report_to=""none"",\n)\ntrainer = Trainer(\n model=model,\n args=args,\n train_dataset=tokenized[""train""],\n eval_dataset=tokenized[""validation""],\n data_collator=collator,\n tokenizer=tokenizer,\n)\n\ntrainer.train()\nAny insights or references to similar cases would be greatly appreciated!
\nThanks in advance.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-16T10:05:35.536Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 8, 'readers_count': 7, 'score': 71.6, 'yours': False, 'topic_id': 166964, 'topic_slug': 'qlora-fine-tuning-is-too-slow-on-llama-based-model-despite-bitsandbytes-optimization', 'display_username': 'Aylin Naebzadeh', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/universitytehran/PersianMind-v1.0', 'internal': False, 'reflection': False, 'title': 'universitytehran/PersianMind-v1.0 · Hugging Face', 'clicks': 0}, {'url': 'https://www.kaggle.com/datasets/zahrarazaghi/parsmap/versions/1', 'internal': False, 'reflection': False, 'title': 'ParsMap | Kaggle', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 60014, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/qlora-fine-tuning-is-too-slow-on-llama-based-model-despite-bitsandbytes-optimization/166964/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 238778, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-16T11:58:54.754Z', 'cooked': '\n\n\n
tf32=True
This wouldn’t work with T4 generation (Maybe Turing) GPUs. Using fp16 will allow you to take advantage of the hardware.
\n\n\n\n
gradient_checkpointing=True,
\ngradient_checkpointing_kwargs={""use_reentrant"": False},
It saves VRAM but slows down the training speed.
\n\n\n\n
target_modules=[""q_proj"",""k_proj"",""v_proj"",""o_proj"",""gate_proj"",""up_proj"",""down_proj""],
As the number of layers to be trained increases, the amount of computation will likely increase, causing the process to slow down.
\nWith shorter sentences, packing=True may be effective. If you want faster trainer, try optimized version.
Thank you!
\nI was able to decrease the time to 23 hours instead of 75 hours!
\nWhich target_modules do you suggest to train?
\nI’ve tried a lot to use SFTTrainer but all the time it raise an error due to versioning and then CUDA out of memory..
I don’t know Kaggle’s etiquette…
\nIs it like this?
pip install -U --no-cache-dir \\\n ""trl==0.18.2"" \\\n ""transformers==4.52.3"" \\\n ""datasets>=2.20.0"" \\\n ""accelerate>=1.2.0"" \\\n ""peft>=0.16.0"" \\\n ""huggingface_hub>=0.23.0"" \\\n ""safetensors>=0.4.3"" \\\n ""bitsandbytes==0.43.1""\npython - <<\'PY\'\nimport IPython; IPython.Application.instance().kernel.do_shutdown(True)\nPY\n\n\nWhich
\ntarget_modulesdo you suggest to train?
target_modules=[""q_proj"",""k_proj"",""v_proj"",""o_proj""],
\nI think many people do this. Is it the attention module? It feels like fine-tuning only that part.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-08-17T11:29:35.101Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.4, 'yours': False, 'topic_id': 166964, 'topic_slug': 'qlora-fine-tuning-is-too-slow-on-llama-based-model-despite-bitsandbytes-optimization', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/qlora-fine-tuning-is-too-slow-on-llama-based-model-despite-bitsandbytes-optimization/166964/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I’m fine-tuning a LLaMA-based model (universitytehran/PersianMind-v1.0) using QLoRA and BitsAndBytes in 4-bit precision. I am working with Kaggle GPU T4, and it takes about 75 hours to be fine-tuned using ParsMap dataset with 40,000 records for training related to converting informal to formal text.
+Here is my code:
base_model_id = ""universitytehran/PersianMind-v1.0""
+compute_dtype = torch.bfloat16 if torch.cuda.get_device_capability(0)[0] >= 8 else torch.float16
+
+print(""Compute dtype:"", compute_dtype)
+def safe_str(x):
+ return """" if x is None or (isinstance(x, float) and np.isnan(x)) else str(x)
+
+df = df_parsmap.copy()
+df = df.dropna(subset=[""inFormalForm"",""formalForm""]) # keep only rows with both sides
+
+def make_text(row):
+ informal = safe_str(row[""inFormalForm""])
+ formal = safe_str(row[""formalForm""])
+ return f""<s><|startoftext|>[Informal]{informal}[Formal]{formal}<|endoftext|>""
+
+df[""text""] = df.apply(make_text, axis=1)
+perm = np.random.permutation(len(df))
+cut = int(0.9*len(df))
+train_df = df.iloc[perm[:cut]].reset_index(drop=True)
+val_df = df.iloc[perm[cut:]].reset_index(drop=True)
+
+ds = DatasetDict({
+ ""train"": Dataset.from_pandas(train_df[[""text""]]),
+ ""validation"": Dataset.from_pandas(val_df[[""text""]]),
+})
+len(ds[""train""]), len(ds[""validation""])
+
+tokenizer = AutoTokenizer.from_pretrained(base_model_id, use_fast=True, trust_remote_code=True)
+
+specials = {
+ ""bos_token"": ""<s>"",
+ ""eos_token"": ""</s>"",
+ ""pad_token"": ""<pad>"",
+}
+
+for k,v in specials.items():
+ if getattr(tokenizer, k, None) != v:
+ tokenizer.add_special_tokens({k: v})
+
+added = tokenizer.add_tokens([""<|startoftext|>"", ""<|endoftext|>"", ""[Informal]"", ""[Formal]"", ""<sep>""], special_tokens=True)
+print(""Added new tokens:"", added)
+
+
+if tokenizer.pad_token is None:
+ tokenizer.pad_token = tokenizer.eos_token
+bnb_config = BitsAndBytesConfig(
+ load_in_4bit=True,
+ bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_use_double_quant=True,
+ bnb_4bit_compute_dtype=compute_dtype,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+ base_model_id,
+ trust_remote_code=True,
+ quantization_config=bnb_config,
+ device_map=""auto"",
+)
+
+model.resize_token_embeddings(len(tokenizer))
+
+model = prepare_model_for_kbit_training(model)
+model.config.use_cache = False
+lora_config = LoraConfig(
+ r=16, lora_alpha=32, lora_dropout=0.1, bias=""none"", task_type=""CAUSAL_LM"",
+ target_modules=[""q_proj"",""k_proj"",""v_proj"",""o_proj"",""gate_proj"",""up_proj"",""down_proj""],
+)
+model = get_peft_model(model, lora_config)
+
+model.gradient_checkpointing_enable()
+
+# quick param report
+trainable = sum(p.numel() for p in model.parameters() if p.requires_grad)
+total = sum(p.numel() for p in model.parameters())
+print(f""Trainable: {trainable:,} / Total: {total:,} ({100*trainable/total:.2f}%)"")
+max_length = 128
+
+def tokenize_batch(batch):
+ return tokenizer(
+ batch[""text""],
+ truncation=True,
+ max_length=max_length,
+ padding=""max_length"",
+ )
+
+tokenized = ds.map(tokenize_batch, batched=True, remove_columns=ds[""train""].column_names)
+collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)
effective_bs = 16
+per_device_train_bs = 2
+per_device_eval_bs = 2
+grad_accum = max(1, effective_bs // per_device_train_bs)
+epochs = 3
+
+args = TrainingArguments(
+ output_dir=""./persianmind-formalizer-lora"",
+ num_train_epochs=epochs,
+ per_device_train_batch_size=per_device_train_bs,
+ per_device_eval_batch_size=per_device_eval_bs,
+ gradient_accumulation_steps=grad_accum,
+ learning_rate=1e-5,
+ warmup_ratio=0.03,
+ lr_scheduler_type=""cosine"",
+ weight_decay=0.0,
+ logging_steps=50,
+
+ eva_strategy=""steps"",
+ eval_steps=2000,
+ save_strategy=""epoch"",
+ save_total_limit=2,
+ load_best_model_at_end=True,
+
+ bf16=(compute_dtype==torch.bfloat16),
+ fp16=(compute_dtype==torch.float16),
+
+ optim=""paged_adamw_8bit"",
+ gradient_checkpointing=True,
+ gradient_checkpointing_kwargs={""use_reentrant"": False},
+
+ dataloader_num_workers=4,
+ dataloader_pin_memory=True,
+ dataloader_persistent_workers=True,
+
+ group_by_length=True,
+ tf32=True,
+ report_to=""none"",
+)
+trainer = Trainer(
+ model=model,
+ args=args,
+ train_dataset=tokenized[""train""],
+ eval_dataset=tokenized[""validation""],
+ data_collator=collator,
+ tokenizer=tokenizer,
+)
+
+trainer.train()
+Any insights or references to similar cases would be greatly appreciated!
+Thanks in advance.
","+++
tf32=True
This wouldn’t work with T4 generation (Maybe Turing) GPUs. Using fp16 will allow you to take advantage of the hardware.
++++
gradient_checkpointing=True,
+gradient_checkpointing_kwargs={""use_reentrant"": False},
It saves VRAM but slows down the training speed.
++++
target_modules=[""q_proj"",""k_proj"",""v_proj"",""o_proj"",""gate_proj"",""up_proj"",""down_proj""],
As the number of layers to be trained increases, the amount of computation will likely increase, causing the process to slow down.
+With shorter sentences, packing=True may be effective. If you want faster trainer, try optimized version.
I have been facing this error and even after checking similar discussions and adding enableXsrfProtection false to my config.toml file, I keep getting this. The upload bar fills up completely but it raises the error afterwards. In some discussions on Streamlit forums people also recommended adding enableCORS = false to the config, which I did but with no result. Tried it in incognito mode but also doesn’t work. Any idea on what might be causing this? If necessary I can provide the files to debug
This issue has existed for quite some time, and there is no known solution. Existing workarounds also do not work.
\nI first asked the AI to summarize the workarounds currently known to address this issue. I will use this as a starting point to explore possible solutions.
\nDo these steps in order.
\nXSRF cookies are restricted inside the Spaces iframe. Streamlit’s uploader then rejects the final POST with 403. (Hugging Face)
\nAdd to your app and check on the deployed Space:
\n\nimport streamlit as st\n\nst.write(""XSRF:"", st.get_option(""server.enableXsrfProtection""))\n\nIt must show False. If it shows True, your config is not loaded. The config file must be at .streamlit/config.toml in the same working directory where streamlit run executes. (Streamlit document)
Create .streamlit/config.toml:
\n[server]\n\nenableXsrfProtection = false\n\n# optional if you test large files:\n\n# maxUploadSize = 400\n\nRedeploy, then re-check step 2. Max upload defaults to 200 MB; increase only if needed. (Streamlit document)
\nStart Streamlit with flags so the setting is guaranteed:
\n\n# Dockerfile (tail)\n\nEXPOSE 8501\n\nCMD streamlit run /app/app.py --server.port=8501 --server.address=0.0.0.0 --server.enableXsrfProtection=false\n\nSpaces’ Streamlit guide shows the SDK and deployment pattern; running on 8501 is standard. (Hugging Face)
\nenableCORS for this error403 on upload is almost always XSRF, not CORS, when embedded or proxied. (Streamlit)
\nThis isolates XSRF from size issues. If large files fail only due to size, raise server.maxUploadSize as in step 3. (Streamlit document)
Azure AD and some proxies strip or alter the _xsrf cookie. Either keep XSRF off or ensure cookies pass through with proper attributes. (GitHub)
If you’re on 1.30.x and still see 403 locally or elsewhere, upgrade to a newer Streamlit release; some users report 1.31.0+ resolved their 403. (Streamlit)
\nA stale client cookie can mismatch after redeploys. Hard refresh after redeploys if you still see 403. (Streamlit)
\nCopy-paste samples you can deploy:
\nA) SDK Space
\n\n# .streamlit/config.toml\n\n[server]\n\nenableXsrfProtection = false\n\n\n# app.py\n\nimport streamlit as st\n\nst.write(""XSRF:"", st.get_option(""server.enableXsrfProtection""))\n\nf = st.file_uploader(""Upload any file"")\n\nif f:\n\ndata = f.getvalue()\n\nst.write({""name"": f.name, ""size_bytes"": len(data)})\n\nB) Docker Space
\n\nFROM python:3.11-slim\n\nRUN pip install --no-cache-dir streamlit==1.38.0\n\nWORKDIR /app\n\nCOPY app.py /app/app.py\n\nEXPOSE 8501\n\nCMD streamlit run /app/app.py --server.port=8501 --server.address=0.0.0.0 --server.enableXsrfProtection=false\n\nThis sequence fixes the 403 in Spaces by ensuring XSRF is actually disabled and verified at runtime. (Hugging Face, Streamlit document)
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-13T15:19:48.547Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 7, 'readers_count': 6, 'score': 56.2, 'yours': False, 'topic_id': 166694, 'topic_slug': 'axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://docs.streamlit.io/develop/api-reference/configuration/config.toml', 'internal': False, 'reflection': False, 'title': 'config.toml - Streamlit Docs', 'clicks': 2}, {'url': 'https://docs.streamlit.io/knowledge-base/deploy/increase-file-uploader-limit-streamlit-cloud', 'internal': False, 'reflection': False, 'title': 'How do I increase the upload limit of st.file_uploader on Streamlit Community Cloud? - Streamlit Docs', 'clicks': 2}, {'url': 'https://huggingface.co/docs/hub/en/spaces-cookie-limitations', 'internal': False, 'reflection': False, 'title': 'Cookie limitations in Spaces', 'clicks': 2}, {'url': 'https://huggingface.co/docs/hub/en/spaces-sdks-streamlit', 'internal': False, 'reflection': False, 'title': 'Streamlit Spaces', 'clicks': 2}, {'url': 'https://discuss.streamlit.io/t/file-uploader-403-error-when-embedding-streamlit-app-in-iframe/24109', 'internal': False, 'reflection': False, 'title': 'File Uploader: 403 error when embedding streamlit app in iframe - Using Streamlit - Streamlit', 'clicks': 0}, {'url': 'https://github.com/streamlit/streamlit/issues/5793', 'internal': False, 'reflection': False, 'title': 'Misconfigured _xsrf cookies · Issue #5793 · streamlit/streamlit · GitHub', 'clicks': 0}, {'url': 'https://discuss.streamlit.io/t/file-upload-fails-with-axioserror-request-failed-with-status-code-403/60945', 'internal': False, 'reflection': False, 'title': 'File upload fails with AxiosError: Request failed with status code 403 - Community Cloud - Streamlit', 'clicks': 0}, {'url': 'https://discuss.streamlit.io/t/file-upload-fails-with-error-request-failed-with-status-code-403/27143?page=4', 'internal': False, 'reflection': False, 'title': 'File upload fails with Error: Request failed with status code 403 - Page 4 - Community Cloud - Streamlit', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit/166694/2', 'reactions': [{'id': 'clap', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 238374, 'name': 'Hugo Torres', 'username': 'HugoFTorres', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugoftorres/{size}/52535_2.png', 'created_at': '2025-08-13T15:29:47.789Z', 'cooked': '\nAdding it to the docker intialization solved the issue, seems like the config was not being read at all. Thanks!
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-13T15:29:47.789Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 1, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 26.0, 'yours': False, 'topic_id': 166694, 'topic_slug': 'axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit', 'display_username': 'Hugo Torres', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101662, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit/166694/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238375, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-13T15:30:40.063Z', 'cooked': 'Great! Congrats.
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-08-13T15:30:40.063Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 166694, 'topic_slug': 'axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 101662, 'username': 'HugoFTorres', 'name': 'Hugo Torres', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugoftorres/{size}/52535_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit/166694/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238443, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-14T03:31:02.193Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-08-14T03:31:02.193Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 166694, 'topic_slug': 'axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/axioserror-request-failed-with-status-code-403-when-uploading-a-file-with-streamlit/166694/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have been facing this error and even after checking similar discussions and adding enableXsrfProtection false to my config.toml file, I keep getting this. The upload bar fills up completely but it raises the error afterwards. In some discussions on Streamlit forums people also recommended adding enableCORS = false to the config, which I did but with no result. Tried it in incognito mode but also doesn’t work. Any idea on what might be causing this? If necessary I can provide the files to debug
This issue has existed for quite some time, and there is no known solution. Existing workarounds also do not work.
+I first asked the AI to summarize the workarounds currently known to address this issue. I will use this as a starting point to explore possible solutions.
+Do these steps in order.
+XSRF cookies are restricted inside the Spaces iframe. Streamlit’s uploader then rejects the final POST with 403. (Hugging Face)
+Add to your app and check on the deployed Space:
+
+import streamlit as st
+
+st.write(""XSRF:"", st.get_option(""server.enableXsrfProtection""))
+
+It must show False. If it shows True, your config is not loaded. The config file must be at .streamlit/config.toml in the same working directory where streamlit run executes. (Streamlit document)
Create .streamlit/config.toml:
+[server]
+
+enableXsrfProtection = false
+
+# optional if you test large files:
+
+# maxUploadSize = 400
+
+Redeploy, then re-check step 2. Max upload defaults to 200 MB; increase only if needed. (Streamlit document)
+Start Streamlit with flags so the setting is guaranteed:
+
+# Dockerfile (tail)
+
+EXPOSE 8501
+
+CMD streamlit run /app/app.py --server.port=8501 --server.address=0.0.0.0 --server.enableXsrfProtection=false
+
+Spaces’ Streamlit guide shows the SDK and deployment pattern; running on 8501 is standard. (Hugging Face)
+enableCORS for this error403 on upload is almost always XSRF, not CORS, when embedded or proxied. (Streamlit)
+This isolates XSRF from size issues. If large files fail only due to size, raise server.maxUploadSize as in step 3. (Streamlit document)
Azure AD and some proxies strip or alter the _xsrf cookie. Either keep XSRF off or ensure cookies pass through with proper attributes. (GitHub)
If you’re on 1.30.x and still see 403 locally or elsewhere, upgrade to a newer Streamlit release; some users report 1.31.0+ resolved their 403. (Streamlit)
+A stale client cookie can mismatch after redeploys. Hard refresh after redeploys if you still see 403. (Streamlit)
+Copy-paste samples you can deploy:
+A) SDK Space
+
+# .streamlit/config.toml
+
+[server]
+
+enableXsrfProtection = false
+
+
+# app.py
+
+import streamlit as st
+
+st.write(""XSRF:"", st.get_option(""server.enableXsrfProtection""))
+
+f = st.file_uploader(""Upload any file"")
+
+if f:
+
+data = f.getvalue()
+
+st.write({""name"": f.name, ""size_bytes"": len(data)})
+
+B) Docker Space
+
+FROM python:3.11-slim
+
+RUN pip install --no-cache-dir streamlit==1.38.0
+
+WORKDIR /app
+
+COPY app.py /app/app.py
+
+EXPOSE 8501
+
+CMD streamlit run /app/app.py --server.port=8501 --server.address=0.0.0.0 --server.enableXsrfProtection=false
+
+This sequence fixes the 403 in Spaces by ensuring XSRF is actually disabled and verified at runtime. (Hugging Face, Streamlit document)
" +"Paper authorship claimed, but still pending",https://discuss.huggingface.co/t/paper-authorship-claimed-but-still-pending/166471,166471,23,2025-08-12 02:56:57.995000+00:00,"[{'id': 237942, 'name': 'Jun Feng', 'username': 'junfeng0288', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/3e96dc/{size}.png', 'created_at': '2025-08-12T02:56:58.053Z', 'cooked': 'I have claimed authorship of this paper, but it has been pending for days now. Please help me with this, thank you!
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-12T02:56:58.053Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 12, 'readers_count': 11, 'score': 57.4, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'Jun Feng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/papers/2508.06009', 'internal': False, 'reflection': False, 'title': 'Paper page - MathReal: We Keep It Real! A Real Scene Benchmark for Evaluating Math Reasoning in Multimodal Large Language Models', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101511, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 237943, 'name': 'Jun Feng', 'username': 'junfeng0288', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/3e96dc/{size}.png', 'created_at': '2025-08-12T03:14:48.471Z', 'cooked': '@meganariley Please help me with this, thank you very much!
', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-12T03:14:48.471Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 12, 'readers_count': 11, 'score': 17.4, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'Jun Feng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101511, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/2', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238229, 'name': 'Jun Feng', 'username': 'junfeng0288', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/3e96dc/{size}.png', 'created_at': '2025-08-13T06:20:36.588Z', 'cooked': '@meganariley @John6666 Please help me with this, thank you very much!
', 'post_number': 3, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-13T06:20:36.588Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 9, 'readers_count': 8, 'score': 41.8, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'Jun Feng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101511, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238239, 'name': 'hysts', 'username': 'hysts', 'avatar_template': '/user_avatar/discuss.huggingface.co/hysts/{size}/32230_2.png', 'created_at': '2025-08-13T06:33:11.045Z', 'cooked': 'Hi @junfeng0288 , sorry for the inconvenience. I’ve reported the issue internally.
', 'post_number': 4, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-13T06:33:11.045Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 31.8, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'hysts', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 7263, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238263, 'name': 'hysts', 'username': 'hysts', 'avatar_template': '/user_avatar/discuss.huggingface.co/hysts/{size}/32230_2.png', 'created_at': '2025-08-13T08:04:48.754Z', 'cooked': '@junfeng0288 Should be fixed now. Thanks for your patience.
', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-13T08:04:48.754Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 36.6, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'hysts', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 7263, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/5', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 238275, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-13T09:21:51.033Z', 'cooked': 'Thank you! hysts.
', 'post_number': 6, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-13T09:21:51.033Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/6', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238373, 'name': 'Jun Feng', 'username': 'junfeng0288', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/3e96dc/{size}.png', 'created_at': '2025-08-13T15:28:29.348Z', 'cooked': 'Thank you very much!
', 'post_number': 7, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-08-13T15:28:29.348Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'Jun Feng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 7263, 'username': 'hysts', 'name': 'hysts', 'avatar_template': '/user_avatar/discuss.huggingface.co/hysts/{size}/32230_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101511, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/7', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238442, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-14T03:28:58.144Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-08-14T03:28:58.144Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 166471, 'topic_slug': 'paper-authorship-claimed-but-still-pending', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/paper-authorship-claimed-but-still-pending/166471/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have claimed authorship of this paper, but it has been pending for days now. Please help me with this, thank you!
+","@junfeng0288 Should be fixed now. Thanks for your patience.
" +ModuleNotFoundError: No module named ‘transformers’,https://discuss.huggingface.co/t/modulenotfounderror-no-module-named-transformers/11609,11609,9,2021-11-11 21:05:23.353000+00:00,"[{'id': 24972, 'name': 'ardo tee', 'username': 'mashedpotatotime', 'avatar_template': '/user_avatar/discuss.huggingface.co/mashedpotatotime/{size}/3103_2.png', 'created_at': '2021-11-11T21:05:23.422Z', 'cooked': 'Hi! I’ve been having trouble getting transformers to work in Spaces.
When tested in my environment using python -c ""from transformers import pipeline; print(pipeline(\'sentiment-analysis\')(\'we love you\'))"", the results show it’s been properly installed. When imported in Colab it works fine too, but whenever deployed to Spaces it always returns the same ModuleNotFound error. Full traceback message:
Traceback:
\nFile ""/home/user/.local/lib/python3.8/site-packages/streamlit/script_runner.py"", line 354, in _run_script\n exec(code, module.__dict__)File ""/home/user/app/app.py"", line 1, in <module>\n from transformers import pipeline\nIt’s a simple test app using transformers and streamlit, - both of which were reinstalled with pip after creating a new venv and reinstalling tensorflow and pytorch. I also tried cleaning, uninstalling, and reinstalling conda based on advice from another forum. No dice.
Currently using:
\nPython 3.9.4
\nTensorflow 2.7.0
\nPyTorch 1.10.0
\nTransformers 4.12.3
\nStreamlit 1.2.0
Any help greatly appreciated! Thanks
it might be due to not having a requirements file. Here is an example of what your spaces app should have - flax-community/image-captioning at main try adding the requirements as they till the environment what packages to load. Hope this helps.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2021-11-12T06:41:54.938Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 198, 'reads': 221, 'readers_count': 220, 'score': 1114.2, 'yours': False, 'topic_id': 11609, 'topic_slug': 'modulenotfounderror-no-module-named-transformers', 'display_username': 'Nikhil', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/flax-community/image-captioning/tree/main', 'internal': False, 'reflection': False, 'title': 'flax-community/image-captioning at main', 'clicks': 2788}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 5}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4732, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/modulenotfounderror-no-module-named-transformers/11609/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 5}], 'current_user_reaction': None, 'reaction_users_count': 5, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 26022, 'name': 'ardo tee', 'username': 'mashedpotatotime', 'avatar_template': '/user_avatar/discuss.huggingface.co/mashedpotatotime/{size}/3103_2.png', 'created_at': '2021-11-19T23:23:39.383Z', 'cooked': 'That worked perfectly. Thank you!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2021-11-19T23:23:39.383Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 137, 'reads': 206, 'readers_count': 205, 'score': 741.2, 'yours': False, 'topic_id': 11609, 'topic_slug': 'modulenotfounderror-no-module-named-transformers', 'display_username': 'ardo tee', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 4732, 'username': 'NDugar', 'name': 'Nikhil', 'avatar_template': '/user_avatar/discuss.huggingface.co/ndugar/{size}/40501_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4950, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/modulenotfounderror-no-module-named-transformers/11609/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 238096, 'name': 'Yue Zhao', 'username': 'Alwaysboy', 'avatar_template': '/user_avatar/discuss.huggingface.co/alwaysboy/{size}/52486_2.png', 'created_at': '2025-08-12T13:40:25.363Z', 'cooked': 'Same issue and solved by this method, thanks!
', 'post_number': 4, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-12T13:40:25.363Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 7, 'readers_count': 6, 'score': 71.4, 'yours': False, 'topic_id': 11609, 'topic_slug': 'modulenotfounderror-no-module-named-transformers', 'display_username': 'Yue Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101586, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/modulenotfounderror-no-module-named-transformers/11609/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi! I’ve been having trouble getting transformers to work in Spaces.
When tested in my environment using python -c ""from transformers import pipeline; print(pipeline('sentiment-analysis')('we love you'))"", the results show it’s been properly installed. When imported in Colab it works fine too, but whenever deployed to Spaces it always returns the same ModuleNotFound error. Full traceback message:
Traceback:
+File ""/home/user/.local/lib/python3.8/site-packages/streamlit/script_runner.py"", line 354, in _run_script
+ exec(code, module.__dict__)File ""/home/user/app/app.py"", line 1, in <module>
+ from transformers import pipeline
+It’s a simple test app using transformers and streamlit, - both of which were reinstalled with pip after creating a new venv and reinstalling tensorflow and pytorch. I also tried cleaning, uninstalling, and reinstalling conda based on advice from another forum. No dice.
Currently using:
+Python 3.9.4
+Tensorflow 2.7.0
+PyTorch 1.10.0
+Transformers 4.12.3
+Streamlit 1.2.0
Any help greatly appreciated! Thanks
it might be due to not having a requirements file. Here is an example of what your spaces app should have - flax-community/image-captioning at main try adding the requirements as they till the environment what packages to load. Hope this helps.
" +The Gradio API by curl doesn’t work,https://discuss.huggingface.co/t/the-gradio-api-by-curl-doesnt-work/166428,166428,5,2025-08-11 17:10:24.724000+00:00,"[{'id': 237880, 'name': 'Dany Gold', 'username': 'GoldDany', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/g/bbce88/{size}.png', 'created_at': '2025-08-11T17:10:24.780Z', 'cooked': 'I try curl from basic guide, but it throws: 405({ “detail”: “Method Not Allowed”}).
\nCurl: curl -X POST https://golddany-didefbackend.hf.space/call/predict -s -H “Content-Type: application/json” -d ‘{
\n“data”: [
\n“Hello!!”
\n]}’
\n| awk -F’""’ ‘{ print $4}’
\n| read EVENT_ID; curl -N https://golddany-didefbackend.hf.space/call/predict/$EVENT_ID
I can get event_id from first request, but second(../$EVENT_ID) always throws: “Connection broken: InvalidChunkLength(got length b’‘, 0 bytes read)”, InvalidChunkLength(got length b’’, 0 bytes read)
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-11T17:15:06.356Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 19, 'reads': 10, 'readers_count': 9, 'score': 107.0, 'yours': False, 'topic_id': 166428, 'topic_slug': 'the-gradio-api-by-curl-doesnt-work', 'display_username': 'Dany Gold', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://golddany-didefbackend.hf.space/call/predict', 'internal': False, 'reflection': False, 'clicks': 1}, {'url': 'https://golddany-didefbackend.hf.space/call/predict/$EVENT_ID', 'internal': False, 'reflection': False, 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101505, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-gradio-api-by-curl-doesnt-work/166428/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 237918, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-11T23:58:55.733Z', 'cooked': 'Hmm, I think the code is written according to the sample. I don’t know what the problem is…
\nI’ll try experimenting a little later.
It worked for some reason… From the server side, it should be the same thing…
\nimport os, requests\n\nSPACE = ""john6666-apitest1.hf.space""\nAPI_NAME = ""predict""\nHF_TOKEN = os.getenv(""HF_TOKEN"", None)\nbase = f""https://{SPACE}""\n\nauth_h = {""Authorization"": f""Bearer {HF_TOKEN}"", ""Content-Type"": ""application/json""} if HF_TOKEN else {}\nr = requests.post(f""{base}/call/{API_NAME}"", headers=auth_h, json={""data"": [""hi""]}, timeout=30)\nr.raise_for_status()\neid = r.json()[""event_id""]\n\nwith requests.get(f""{base}/call/{API_NAME}/{eid}"", headers={""Authorization"": f""Bearer {HF_TOKEN}"", ""Accept"": ""text/event-stream""}, stream=True, timeout=300) as resp:\n for line in resp.iter_lines(decode_unicode=True):\n if line:\n print(line) # data: [[0.03394877910614014, -0.005614369176328182, -0.0012183655053377151, 0.015974245965480804,...\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-12T13:32:56.414Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 166428, 'topic_slug': 'the-gradio-api-by-curl-doesnt-work', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/the-gradio-api-by-curl-doesnt-work/166428/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I try curl from basic guide, but it throws: 405({ “detail”: “Method Not Allowed”}).
+Curl: curl -X POST https://golddany-didefbackend.hf.space/call/predict -s -H “Content-Type: application/json” -d ‘{
+“data”: [
+“Hello!!”
+]}’
+| awk -F’""’ ‘{ print $4}’
+| read EVENT_ID; curl -N https://golddany-didefbackend.hf.space/call/predict/$EVENT_ID
I can get event_id from first request, but second(../$EVENT_ID) always throws: “Connection broken: InvalidChunkLength(got length b’‘, 0 bytes read)”, InvalidChunkLength(got length b’’, 0 bytes read)
","It worked for some reason… From the server side, it should be the same thing…
+import os, requests
+
+SPACE = ""john6666-apitest1.hf.space""
+API_NAME = ""predict""
+HF_TOKEN = os.getenv(""HF_TOKEN"", None)
+base = f""https://{SPACE}""
+
+auth_h = {""Authorization"": f""Bearer {HF_TOKEN}"", ""Content-Type"": ""application/json""} if HF_TOKEN else {}
+r = requests.post(f""{base}/call/{API_NAME}"", headers=auth_h, json={""data"": [""hi""]}, timeout=30)
+r.raise_for_status()
+eid = r.json()[""event_id""]
+
+with requests.get(f""{base}/call/{API_NAME}/{eid}"", headers={""Authorization"": f""Bearer {HF_TOKEN}"", ""Accept"": ""text/event-stream""}, stream=True, timeout=300) as resp:
+ for line in resp.iter_lines(decode_unicode=True):
+ if line:
+ print(line) # data: [[0.03394877910614014, -0.005614369176328182, -0.0012183655053377151, 0.015974245965480804,...
+the gradio throws error: Traceback (most recent call last):
\nFile “C:\\Users\\danya\\PycharmProjects\\DiDefBackend\\DiDef\\SentenceTransformer.py”, line 45, in
\nclient = Client(
\nFile “C:\\Users\\danya\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\gradio_client\\client.py”, line 171, in init
\nself._info = self._get_api_info()
\nFile “C:\\Users\\danya\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\gradio_client\\client.py”, line 564, in get_api_info
\ninfo = r.json()
\nFile “C:\\Users\\danya\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\httpx_models.py”, line 764, in json
\nreturn jsonlib.loads(self.content, **kwargs)
\nFile ""C:\\Users\\danya\\AppData\\Local\\Programs\\Python\\Python39\\lib\\json_init.py"", line 346, in loads
\nreturn _default_decoder.decode(s)
\nFile “C:\\Users\\danya\\AppData\\Local\\Programs\\Python\\Python39\\lib\\json\\decoder.py”, line 337, in decode
\nobj, end = self.raw_decode(s, idx=_w(s, 0).end())
\nFile “C:\\Users\\danya\\AppData\\Local\\Programs\\Python\\Python39\\lib\\json\\decoder.py”, line 355, in raw_decode
\nraise JSONDecodeError(“Expecting value”, s, err.value) from None
\njson.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
why? My code is very simple:
\nfrom gradio_client import Client
\nclient = Client(
\nsrc = “GoldDany/DiDefBackend”, #my Space is public
\n)
\nresult = client.predict(
\ntext=“Hello!!”,
\napi_name=“/predict”,
\n)
\nprint(result)
\n\nPython39
\n
I think this is probably the culprit this time.
\nGradio 5 only works with Python 3.10 or later on both the server and client, so I think the error is occurring because the versions are different between the client and server.
\nI don’t know if this error can be potentially resolved…
The simplest solution is to use Python 3.10 or later.
# pip install -U gradio_client (in Python 3.9 environment)\nimport subprocess\nsubprocess.run(""pip show gradio_client"", shell=True) # Version: 1.3.0 (Release date: 2024.08.08)\nfrom gradio_client import Client\n\nclient = Client(src=""John6666/apitest1"") # Gradio 4.41.0\nresult = client.predict(text=""Hello!!"", api_name=""/predict"")\nprint(result) # [0.010964062064886093, 0.02713009901344776, -0.024556249380111694, 0.01713254489004612, 0.04088324308395386, -0.005583592690527439, 0.015990763902664185,...\n\nclient = Client(src=""GoldDany/DiDefBackend"") # Gradio 5.42.0\nresult = client.predict(text=""Hello!!"", api_name=""/predict"")\nprint(result) # error\nThanks) But I may have to use an even lower version python, because integrating it . But downgrading the version of Gradio works))
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-12T02:25:10.323Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 166407, 'topic_slug': 'the-gradio-api-is-not-working', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/the-gradio-api-is-not-working/166407/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","the gradio throws error: Traceback (most recent call last):
+File “C:\Users\danya\PycharmProjects\DiDefBackend\DiDef\SentenceTransformer.py”, line 45, in
+client = Client(
+File “C:\Users\danya\AppData\Local\Programs\Python\Python39\lib\site-packages\gradio_client\client.py”, line 171, in init
+self._info = self._get_api_info()
+File “C:\Users\danya\AppData\Local\Programs\Python\Python39\lib\site-packages\gradio_client\client.py”, line 564, in get_api_info
+info = r.json()
+File “C:\Users\danya\AppData\Local\Programs\Python\Python39\lib\site-packages\httpx_models.py”, line 764, in json
+return jsonlib.loads(self.content, **kwargs)
+File ""C:\Users\danya\AppData\Local\Programs\Python\Python39\lib\json_init.py"", line 346, in loads
+return _default_decoder.decode(s)
+File “C:\Users\danya\AppData\Local\Programs\Python\Python39\lib\json\decoder.py”, line 337, in decode
+obj, end = self.raw_decode(s, idx=_w(s, 0).end())
+File “C:\Users\danya\AppData\Local\Programs\Python\Python39\lib\json\decoder.py”, line 355, in raw_decode
+raise JSONDecodeError(“Expecting value”, s, err.value) from None
+json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
why? My code is very simple:
+from gradio_client import Client
+client = Client(
+src = “GoldDany/DiDefBackend”, #my Space is public
+)
+result = client.predict(
+text=“Hello!!”,
+api_name=“/predict”,
+)
+print(result)
++Python39
+
I think this is probably the culprit this time.
+Gradio 5 only works with Python 3.10 or later on both the server and client, so I think the error is occurring because the versions are different between the client and server.
+I don’t know if this error can be potentially resolved…
The simplest solution is to use Python 3.10 or later.
# pip install -U gradio_client (in Python 3.9 environment)
+import subprocess
+subprocess.run(""pip show gradio_client"", shell=True) # Version: 1.3.0 (Release date: 2024.08.08)
+from gradio_client import Client
+
+client = Client(src=""John6666/apitest1"") # Gradio 4.41.0
+result = client.predict(text=""Hello!!"", api_name=""/predict"")
+print(result) # [0.010964062064886093, 0.02713009901344776, -0.024556249380111694, 0.01713254489004612, 0.04088324308395386, -0.005583592690527439, 0.015990763902664185,...
+
+client = Client(src=""GoldDany/DiDefBackend"") # Gradio 5.42.0
+result = client.predict(text=""Hello!!"", api_name=""/predict"")
+print(result) # error
+Hey there !
I am contributing to the translation project for smolagents on the Hugging Face GitHub repository, translating from English to Spanish.
\nHowever, when I try to preview the English documentation (or any other language) using the command
\ndoc-builder preview smolagents docs/source/en, I encounter 404 errors on the main index section, which prevents me from properly previewing the documentation locally (on Mac).
Attached are screenshots illustrating the issue. I would appreciate any guidance on how to resolve this. Thanks in advance for your help!
\nP.S. I also checked the post on Error with Doc-Builder: Error 404 on Section Pages in Doc-Builder Preview , but it didn’t help.
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-09T21:13:46.009Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 6, 'readers_count': 5, 'score': 51.2, 'yours': False, 'topic_id': 166230, 'topic_slug': 'error-with-doc-builder-in-smolagents-documentation-notfound-error', 'display_username': 'David Arias', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/error-with-doc-builder-error-404-on-section-pages-in-doc-builder-preview/68379', 'internal': True, 'reflection': False, 'title': 'Error with Doc-Builder: Error 404 on Section Pages in Doc-Builder Preview', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 74180, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-with-doc-builder-in-smolagents-documentation-notfound-error/166230/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 237545, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-10T00:16:13.835Z', 'cooked': 'There seems to be a version mismatch in the JavaScript version of DocBuilder…
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-10T00:16:13.835Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 21.2, 'yours': False, 'topic_id': 166230, 'topic_slug': 'error-with-doc-builder-in-smolagents-documentation-notfound-error', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/doc-builder/issues/502', 'internal': False, 'reflection': False, 'title': 'NotFound [Error]: Not found: / · Issue #502 · huggingface/doc-builder · GitHub', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-with-doc-builder-in-smolagents-documentation-notfound-error/166230/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237564, 'name': 'David Arias', 'username': 'beta3', 'avatar_template': '/user_avatar/discuss.huggingface.co/beta3/{size}/36181_2.png', 'created_at': '2025-08-10T03:02:16.508Z', 'cooked': 'Thank you ! I tried both version 0.6.0.dev0 and commit 3de0a0e ( GitHub - huggingface/doc-builder at 3de0a0e9f824fc50e78c873732ef4a4ebaeb005b ), but neither worked for me. However, I found a possible temporary workaround to test the documentation locally.
Steps:
\nClone the main repository you want to work with using:
\ngit clone https://github.com/huggingface/smolagents.git
Inside the main folder, run the following commands:
\npip install -e .\npip install watchdog\ngit clone https://github.com/huggingface/doc-builder.git\ncd doc-builder\npip install -e .\ncd ..\nIn the _toctree.yml file (inside the docs/en folder) , change the values on lines 3 and 4 from:
local: index\ntitle: Introduction\nto
\nlocal: index1\ntitle: Introduction1\nand save the file
\nChange the name of the index file from index.md to index1.md
Start the server by running:
\ndoc-builder preview smolagents docs/source/en/
Note: Don’t forget to change the values in _toctree.yml back before pushing your changes to avoid any issues. You can also preview the docs after opening a PR.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-10T16:01:49.037Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 166230, 'topic_slug': 'error-with-doc-builder-in-smolagents-documentation-notfound-error', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/error-with-doc-builder-in-smolagents-documentation-notfound-error/166230/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hey there !
I am contributing to the translation project for smolagents on the Hugging Face GitHub repository, translating from English to Spanish.
+However, when I try to preview the English documentation (or any other language) using the command
+doc-builder preview smolagents docs/source/en, I encounter 404 errors on the main index section, which prevents me from properly previewing the documentation locally (on Mac).
Attached are screenshots illustrating the issue. I would appreciate any guidance on how to resolve this. Thanks in advance for your help!
+P.S. I also checked the post on Error with Doc-Builder: Error 404 on Section Pages in Doc-Builder Preview , but it didn’t help.
+","Thank you ! I tried both version 0.6.0.dev0 and commit 3de0a0e ( GitHub - huggingface/doc-builder at 3de0a0e9f824fc50e78c873732ef4a4ebaeb005b ), but neither worked for me. However, I found a possible temporary workaround to test the documentation locally.
Steps:
+Clone the main repository you want to work with using:
+git clone https://github.com/huggingface/smolagents.git
Inside the main folder, run the following commands:
+pip install -e .
+pip install watchdog
+git clone https://github.com/huggingface/doc-builder.git
+cd doc-builder
+pip install -e .
+cd ..
+In the _toctree.yml file (inside the docs/en folder) , change the values on lines 3 and 4 from:
local: index
+title: Introduction
+to
+local: index1
+title: Introduction1
+and save the file
+Change the name of the index file from index.md to index1.md
Start the server by running:
+doc-builder preview smolagents docs/source/en/
Note: Don’t forget to change the values in _toctree.yml back before pushing your changes to avoid any issues. You can also preview the docs after opening a PR.
Hi everyone!
\nI successfully fine-tuned the meta-llama/Llama-3.1-8B-Instruct model using the dataset G-reen/TheatreLM-v2.1-Characters.
\nThe training was done using LLaMA-Factory, since that was the only method that worked for me.
The training itself went fine. But now I’m stuck with a problem.
\n I don’t understand how to merge the base model and the fine-tuned files into a single
.gguf file so I can use it in LM Studio.
Here’s how my files are organized:
\n Fine-tuned files (LoRA output):
\nD:\\IA\\LLaMA-Factory\\saves\\Llama-3.1-8B\\lora\\train_2025-05-24-18-39-59
Base model:
\nD:\\IA\\LLaMA-Factory\\models\\Llama-3.1-8B
I’ve tried different ways but nothing worked so far.
\nIf anyone can explain how to properly combine these into a .gguf file — I would really appreciate the help!
Thanks in advance!
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-25T09:48:43.119Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 111, 'reads': 9, 'readers_count': 8, 'score': 566.8, 'yours': False, 'topic_id': 156692, 'topic_slug': 'how-to-merge-fine-tuned-llama-3-1-8b-via-llama-factory-into-a-single-gguf-for-lm-studio', 'display_username': 'fsdf', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct', 'internal': False, 'reflection': False, 'title': 'meta-llama/Llama-3.1-8B-Instruct · Hugging Face', 'clicks': 1}, {'url': 'https://huggingface.co/datasets/G-reen/TheatreLM-v2.1-Characters', 'internal': False, 'reflection': False, 'title': 'G-reen/TheatreLM-v2.1-Characters · Datasets at Hugging Face', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95038, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-merge-fine-tuned-llama-3-1-8b-via-llama-factory-into-a-single-gguf-for-lm-studio/156692/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 223932, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-25T10:41:08.007Z', 'cooked': 'Maybe similar case?
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-25T10:41:08.007Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 1.6, 'yours': False, 'topic_id': 156692, 'topic_slug': 'how-to-merge-fine-tuned-llama-3-1-8b-via-llama-factory-into-a-single-gguf-for-lm-studio', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/leafspark/Meta-Llama-3.1-405B-Instruct-GGUF/discussions/2', 'internal': False, 'reflection': False, 'title': 'leafspark/Meta-Llama-3.1-405B-Instruct-GGUF · how to merge all 8 split gguf files', 'clicks': 30}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-merge-fine-tuned-llama-3-1-8b-via-llama-factory-into-a-single-gguf-for-lm-studio/156692/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 237642, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-10T11:40:38.252Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-10T11:40:38.252Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 156692, 'topic_slug': 'how-to-merge-fine-tuned-llama-3-1-8b-via-llama-factory-into-a-single-gguf-for-lm-studio', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-merge-fine-tuned-llama-3-1-8b-via-llama-factory-into-a-single-gguf-for-lm-studio/156692/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone!
+I successfully fine-tuned the meta-llama/Llama-3.1-8B-Instruct model using the dataset G-reen/TheatreLM-v2.1-Characters.
+The training was done using LLaMA-Factory, since that was the only method that worked for me.
The training itself went fine. But now I’m stuck with a problem.
+ I don’t understand how to merge the base model and the fine-tuned files into a single
.gguf file so I can use it in LM Studio.
Here’s how my files are organized:
+ Fine-tuned files (LoRA output):
+D:\IA\LLaMA-Factory\saves\Llama-3.1-8B\lora\train_2025-05-24-18-39-59
Base model:
+D:\IA\LLaMA-Factory\models\Llama-3.1-8B
I’ve tried different ways but nothing worked so far.
+If anyone can explain how to properly combine these into a .gguf file — I would really appreciate the help!
Thanks in advance!
+","Maybe similar case?
+" +To calibrate or not to calibrate for ranking?,https://discuss.huggingface.co/t/to-calibrate-or-not-to-calibrate-for-ranking/166132,166132,5,2025-08-08 14:39:07.163000+00:00,"[{'id': 237362, 'name': 'John do', 'username': 'JPFrancoia', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/dbc845/{size}.png', 'created_at': '2025-08-08T14:39:07.224Z', 'cooked': 'Hi,
\nI made and fine-tuned a binary text classifier with ModernBERT. My objective is to rank by (relevance) news articles coming from my RSS feeds. I labelled some “good” articles (interesting and relevant to me) and some “bad” articles (irrelevant to me) and fine-tuned the classifier on them.
\nWith this trained classifier, I’m trying to assign a relevance score to any unread article. Ultimately, the articles with the highest score will land at the top of my reading queue, and I can prioritise them. The only thing I really care about is the ranking.
\nBut here is the problem: I trained this classifier once, but I perform inference every hour, to make sure the new unread articles get evaluated. So I need a scoring technique that is consistent across inference runs. For example, article A gets scored at 8am (in a batch of 100 articles) and get a score of 42. If it gets re-evaluated at 2pm in another batch of 200 articles, it needs to get a score of 42 again. Otherwise, the ranking will be completely unreliable.
\nUnfortunately my maths skills don’t allow me to answer this question myself:
\nFor the sigmoid part, I have something like that:
\ninputs = tokenizer(\n batch_texts,\n padding=True,\n truncation=True,\n max_length=MAX_LENGTH,\n return_tensors=""pt"",\n)\npreds = model(**inputs).logits\nprobs = torch.sigmoid(preds[:, 1]).cpu().numpy()\nI could also do this to calibrate the probabilities:
\nlogit_diff = all_logits[:, 1] - all_logits[:, 0]\ncalibrator = LogisticRegression()\ncalibrator.fit(logit_diff.reshape(-1, 1), true_labels)\nBut I don’t know if I should or shouldn’t calibrate…
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-08T14:39:07.224Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 14, 'reads': 6, 'readers_count': 5, 'score': 86.2, 'yours': False, 'topic_id': 166132, 'topic_slug': 'to-calibrate-or-not-to-calibrate-for-ranking', 'display_username': 'John do', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98130, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/to-calibrate-or-not-to-calibrate-for-ranking/166132/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 237435, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-09T00:09:06.247Z', 'cooked': 'My mathematical knowledge is hopeless😭, but I don’t think calibration is necessary for the rankings…
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-09T00:09:06.247Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 11.0, 'yours': False, 'topic_id': 166132, 'topic_slug': 'to-calibrate-or-not-to-calibrate-for-ranking', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://scikit-learn.org/stable/modules/calibration.html', 'internal': False, 'reflection': False, 'title': '1.16. Probability calibration — scikit-learn 1.7.1 documentation', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/to-calibrate-or-not-to-calibrate-for-ranking/166132/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 237470, 'name': 'John do', 'username': 'JPFrancoia', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/dbc845/{size}.png', 'created_at': '2025-08-09T10:39:56.284Z', 'cooked': 'It is generally expected that calibration does not affect ranking
\n
Thank you very much!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-08-09T10:39:56.284Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 166132, 'topic_slug': 'to-calibrate-or-not-to-calibrate-for-ranking', 'display_username': 'John do', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98130, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/to-calibrate-or-not-to-calibrate-for-ranking/166132/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237532, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-09T22:40:51.541Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-08-09T22:40:51.541Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 166132, 'topic_slug': 'to-calibrate-or-not-to-calibrate-for-ranking', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/to-calibrate-or-not-to-calibrate-for-ranking/166132/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I made and fine-tuned a binary text classifier with ModernBERT. My objective is to rank by (relevance) news articles coming from my RSS feeds. I labelled some “good” articles (interesting and relevant to me) and some “bad” articles (irrelevant to me) and fine-tuned the classifier on them.
+With this trained classifier, I’m trying to assign a relevance score to any unread article. Ultimately, the articles with the highest score will land at the top of my reading queue, and I can prioritise them. The only thing I really care about is the ranking.
+But here is the problem: I trained this classifier once, but I perform inference every hour, to make sure the new unread articles get evaluated. So I need a scoring technique that is consistent across inference runs. For example, article A gets scored at 8am (in a batch of 100 articles) and get a score of 42. If it gets re-evaluated at 2pm in another batch of 200 articles, it needs to get a score of 42 again. Otherwise, the ranking will be completely unreliable.
+Unfortunately my maths skills don’t allow me to answer this question myself:
+For the sigmoid part, I have something like that:
+inputs = tokenizer(
+ batch_texts,
+ padding=True,
+ truncation=True,
+ max_length=MAX_LENGTH,
+ return_tensors=""pt"",
+)
+preds = model(**inputs).logits
+probs = torch.sigmoid(preds[:, 1]).cpu().numpy()
+I could also do this to calibrate the probabilities:
+logit_diff = all_logits[:, 1] - all_logits[:, 0]
+calibrator = LogisticRegression()
+calibrator.fit(logit_diff.reshape(-1, 1), true_labels)
+But I don’t know if I should or shouldn’t calibrate…
","My mathematical knowledge is hopeless😭, but I don’t think calibration is necessary for the rankings…
+ ++" +The Best Approach for Weighted Multilabel Classification,https://discuss.huggingface.co/t/the-best-approach-for-weighted-multilabel-classification/137121,137121,9,2025-01-24 07:13:46.641000+00:00,"[{'id': 197515, 'name': 'Aylin Naebzadeh', 'username': 'AylinNaebzadeh', 'avatar_template': '/user_avatar/discuss.huggingface.co/aylinnaebzadeh/{size}/52343_2.png', 'created_at': '2025-01-24T07:13:46.720Z', 'cooked': 'It is generally expected that calibration does not affect ranking
+
Hello.
\nI have a task in which there are 6 different labels for each record, and every label can have a value from 0 to 3. The dataset is so imbalanced.
\n| text | \nlabel_1 | \nlabel_2 | \nlabel_3 | \nlabel_4 | \nlabel_5 | \nlabel_6 | \n
|---|---|---|---|---|---|---|
| … | \n0 | \n1 | \n0 | \n2 | \n0 | \n0 | \n
| … | \n0 | \n0 | \n0 | \n0 | \n0 | \n0 | \n
| … | \n2 | \n0 | \n0 | \n0 | \n0 | \n3 | \n
I want to solve this task using transformers. Should I set the num_labels equal to 24 while initializing the transformer?
num_labels = 6 # Number of labels\nclasses_per_label = 4 # Number of intensity levels per label (0, 1, 2, 3)\ntotal_classes = num_labels * classes_per_label\n\nmodel = AutoModelForSequenceClassification.from_pretrained(model_name,\n problem_type=""multi_label_classification"",\n ignore_mismatched_sizes=True,\n num_labels=total_classes)\nIn addition, what are best practices for 1. creating a Dataset object from torch.utils.data.Dataset module, 2. defining a loss function, and 3. defining thresholds while predicting and evaluating the labels?
Here is my current code:
\ndef encode_data(df, tokenizer, label_columns):\n encodings = tokenizer(list(df[\'text\']), padding=True, truncation=True, max_length=128)\n labels = df[label_columns].values\n return encodings, labels\n\nclass WeightedMultiLabelDataset(torch.utils.data.Dataset):\n def __init__(self, encodings, labels):\n self.encodings = encodings\n self.labels = torch.tensor(labels, dtype=torch.long)\n\n def __len__(self):\n return len(self.labels)\n\n def __getitem__(self, idx):\n item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}\n item[\'labels\'] = self.labels[idx]\n return item\n\n# Prepare datasets\ntrain_encodings, train_labels = encode_data(train_df, tokenizer, label_columns)\ndev_encodings, dev_labels = encode_data(dev_df, tokenizer, label_columns)\n\ntrain_dataset = WeightedMultiLabelDataset(train_encodings, train_labels)\ndev_dataset = WeightedMultiLabelDataset(dev_encodings, dev_labels)\nfrom sklearn.metrics import classification_report, average_precision_score\n\ndef compute_metrics(pred):\n logits, labels = pred\n \n logits = logits.reshape(-1, classes_per_label)\n probabilities = torch.softmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()\n predictions = torch.argmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()\n labels = labels.reshape(-1, num_labels).numpy()\n\n auprc_per_label = []\n for i in range(num_labels):\n auprc = average_precision_score(labels[:, i], probabilities[:, i])\n auprc_per_label.append(auprc)\n \n mean_auprc = sum(auprc_per_label) / len(auprc_per_label)\n\n report = classification_report(labels, predictions, target_names=label_columns, zero_division=0)\n print(report)\n\n return {\n \'mean_auprc\': mean_auprc,\n \'auprc_per_label\': auprc_per_label,\n }\nThank you!
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-01-24T07:18:42.126Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 35, 'reads': 10, 'readers_count': 9, 'score': 192.0, 'yours': False, 'topic_id': 137121, 'topic_slug': 'the-best-approach-for-weighted-multilabel-classification', 'display_username': 'Aylin Naebzadeh', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 60014, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-best-approach-for-weighted-multilabel-classification/137121/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 197594, 'name': 'Alan turner', 'username': 'Alanturner2', 'avatar_template': '/user_avatar/discuss.huggingface.co/alanturner2/{size}/37542_2.png', 'created_at': '2025-01-24T14:01:36.482Z', 'cooked': 'Hi there, I read your question and can see you’re working on an interesting multi-label classification task. Let me help clarify your doubts and provide some guidance on best practices.
\nFirst, regarding num_labels, setting it equal to 24 (6 labels × 4 intensity levels) is incorrect. For your case, each label is independent and can take one of four values (0, 1, 2, 3). You should set num_labels = 6 when initializing your transformer. This is because you’re solving a multi-label classification problem, where each label is treated as a separate classification task with its own probabilities.
For the rest of your queries, here are my suggestions:
\nDataset ObjectYour current implementation of the WeightedMultiLabelDataset is good, but since your task deals with integer values (0–3) for each label, you need to ensure the labels are properly encoded. You should consider using torch.float instead of torch.long if you’re working with one-hot or probabilities for evaluation.
Also, verify that your tokenizer outputs include all necessary fields like input_ids, attention_mask, and optionally token_type_ids.
For this task, you can use torch.nn.CrossEntropyLoss for each label since your labels are categorical with four classes. Since your dataset is imbalanced, consider using class weights to handle the imbalance effectively. Here’s an example:
loss_fn = torch.nn.CrossEntropyLoss(weight=class_weights) \nYou can calculate class_weights using the frequency of each class in your dataset.
During prediction, you can use torch.softmax to get the probabilities for each intensity level. To evaluate, you can use torch.argmax to select the most probable intensity level for each label. No additional thresholds are necessary since your task involves classification rather than binary decisions.
Here’s how you can adjust your code:
\nlogits = logits.reshape(-1, classes_per_label)\nprobabilities = torch.softmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()\npredictions = torch.argmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()\nWeightedRandomSampler during training to address class imbalance.Here’s a slightly modified version of your dataset class:
\nclass WeightedMultiLabelDataset(torch.utils.data.Dataset):\n def __init__(self, encodings, labels):\n self.encodings = encodings\n self.labels = torch.tensor(labels, dtype=torch.float) # Use float if needed for evaluation\n\n def __len__(self):\n return len(self.labels)\n\n def __getitem__(self, idx):\n item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}\n item[\'labels\'] = self.labels[idx]\n return item\nYour approach is solid! By following these adjustments, you should be able to handle the multi-label classification task effectively. Let me know if you need further clarification or assistance. Good luck!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-09T15:56:12.152Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 137121, 'topic_slug': 'the-best-approach-for-weighted-multilabel-classification', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/the-best-approach-for-weighted-multilabel-classification/137121/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello.
+I have a task in which there are 6 different labels for each record, and every label can have a value from 0 to 3. The dataset is so imbalanced.
+| text | +label_1 | +label_2 | +label_3 | +label_4 | +label_5 | +label_6 | +
|---|---|---|---|---|---|---|
| … | +0 | +1 | +0 | +2 | +0 | +0 | +
| … | +0 | +0 | +0 | +0 | +0 | +0 | +
| … | +2 | +0 | +0 | +0 | +0 | +3 | +
I want to solve this task using transformers. Should I set the num_labels equal to 24 while initializing the transformer?
num_labels = 6 # Number of labels
+classes_per_label = 4 # Number of intensity levels per label (0, 1, 2, 3)
+total_classes = num_labels * classes_per_label
+
+model = AutoModelForSequenceClassification.from_pretrained(model_name,
+ problem_type=""multi_label_classification"",
+ ignore_mismatched_sizes=True,
+ num_labels=total_classes)
+In addition, what are best practices for 1. creating a Dataset object from torch.utils.data.Dataset module, 2. defining a loss function, and 3. defining thresholds while predicting and evaluating the labels?
Here is my current code:
+def encode_data(df, tokenizer, label_columns):
+ encodings = tokenizer(list(df['text']), padding=True, truncation=True, max_length=128)
+ labels = df[label_columns].values
+ return encodings, labels
+
+class WeightedMultiLabelDataset(torch.utils.data.Dataset):
+ def __init__(self, encodings, labels):
+ self.encodings = encodings
+ self.labels = torch.tensor(labels, dtype=torch.long)
+
+ def __len__(self):
+ return len(self.labels)
+
+ def __getitem__(self, idx):
+ item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
+ item['labels'] = self.labels[idx]
+ return item
+
+# Prepare datasets
+train_encodings, train_labels = encode_data(train_df, tokenizer, label_columns)
+dev_encodings, dev_labels = encode_data(dev_df, tokenizer, label_columns)
+
+train_dataset = WeightedMultiLabelDataset(train_encodings, train_labels)
+dev_dataset = WeightedMultiLabelDataset(dev_encodings, dev_labels)
+from sklearn.metrics import classification_report, average_precision_score
+
+def compute_metrics(pred):
+ logits, labels = pred
+
+ logits = logits.reshape(-1, classes_per_label)
+ probabilities = torch.softmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()
+ predictions = torch.argmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()
+ labels = labels.reshape(-1, num_labels).numpy()
+
+ auprc_per_label = []
+ for i in range(num_labels):
+ auprc = average_precision_score(labels[:, i], probabilities[:, i])
+ auprc_per_label.append(auprc)
+
+ mean_auprc = sum(auprc_per_label) / len(auprc_per_label)
+
+ report = classification_report(labels, predictions, target_names=label_columns, zero_division=0)
+ print(report)
+
+ return {
+ 'mean_auprc': mean_auprc,
+ 'auprc_per_label': auprc_per_label,
+ }
+Thank you!
","Hi there, I read your question and can see you’re working on an interesting multi-label classification task. Let me help clarify your doubts and provide some guidance on best practices.
+First, regarding num_labels, setting it equal to 24 (6 labels × 4 intensity levels) is incorrect. For your case, each label is independent and can take one of four values (0, 1, 2, 3). You should set num_labels = 6 when initializing your transformer. This is because you’re solving a multi-label classification problem, where each label is treated as a separate classification task with its own probabilities.
For the rest of your queries, here are my suggestions:
+Dataset ObjectYour current implementation of the WeightedMultiLabelDataset is good, but since your task deals with integer values (0–3) for each label, you need to ensure the labels are properly encoded. You should consider using torch.float instead of torch.long if you’re working with one-hot or probabilities for evaluation.
Also, verify that your tokenizer outputs include all necessary fields like input_ids, attention_mask, and optionally token_type_ids.
For this task, you can use torch.nn.CrossEntropyLoss for each label since your labels are categorical with four classes. Since your dataset is imbalanced, consider using class weights to handle the imbalance effectively. Here’s an example:
loss_fn = torch.nn.CrossEntropyLoss(weight=class_weights)
+You can calculate class_weights using the frequency of each class in your dataset.
During prediction, you can use torch.softmax to get the probabilities for each intensity level. To evaluate, you can use torch.argmax to select the most probable intensity level for each label. No additional thresholds are necessary since your task involves classification rather than binary decisions.
Here’s how you can adjust your code:
+logits = logits.reshape(-1, classes_per_label)
+probabilities = torch.softmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()
+predictions = torch.argmax(torch.tensor(logits), axis=1).view(-1, num_labels).numpy()
+WeightedRandomSampler during training to address class imbalance.Here’s a slightly modified version of your dataset class:
+class WeightedMultiLabelDataset(torch.utils.data.Dataset):
+ def __init__(self, encodings, labels):
+ self.encodings = encodings
+ self.labels = torch.tensor(labels, dtype=torch.float) # Use float if needed for evaluation
+
+ def __len__(self):
+ return len(self.labels)
+
+ def __getitem__(self, idx):
+ item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
+ item['labels'] = self.labels[idx]
+ return item
+Your approach is solid! By following these adjustments, you should be able to handle the multi-label classification task effectively. Let me know if you need further clarification or assistance. Good luck!
Hello,
\nI am currently trying to generate a lot of embeddings as part of a research project for my company.
\nWe have a team account setup as well as a valid billing method, and a token associated to our company in order to perform API calls.
\nI’m using Qwen3-Embeddings-8B ( Qwen/Qwen3-Embedding-8B · Hugging Face )
\nI can call it and get some embeddings, but after around 3000 or so embeddings I get hit with a limit and receive a 402 “Payment Required” exception. This surprised me since we do have a billing method.
\nThen I looked into it a bit more and saw that “Inference Usage” has a max limit of $0 per month unless you have a team/entreprise account. So that means that you can’t pay per usage at all as a company until you set that up? Am I understading this correctly?
', 'post_number': 1, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T08:20:45.901Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 6, 'readers_count': 5, 'score': 46.2, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'Luca Rizzello', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/Qwen/Qwen3-Embedding-8B?text=hi&inference_api=true&inference_provider=nebius&language=python&client=huggingface_hub', 'internal': False, 'reflection': False, 'title': 'Qwen/Qwen3-Embedding-8B · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101215, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 237116, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-07T10:28:10.908Z', 'cooked': 'I believe that a Pro, Teams, or Enterprise subscription is required for PAYG billing for Inference Provider (at least for now). It would be best to check with Hugging Face support to be certain. billing@huggingface.co
', 'post_number': 2, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T10:28:10.908Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/hugging-face-payment-error-402-youve-exceeded-monthly-quota/144968/20', 'internal': True, 'reflection': False, 'title': ""Hugging Face Payment Error 402 & You've Exceeded Monthly Quota"", 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237119, 'name': 'Luca Rizzello', 'username': 'lrizzellotaskbase', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/l/da6949/{size}.png', 'created_at': '2025-08-07T10:41:10.791Z', 'cooked': 'Thanks for the reply. I’ll mail HF directly
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T10:41:10.791Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 26.0, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'Luca Rizzello', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101215, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237161, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-08-07T14:34:33.046Z', 'cooked': 'Hi @lrizzellotaskbase Thanks for posting! Upgrading your org to Team or Enterprise for Inference Providers usage comes with many perks:
\nYour organization has a pool of $2 of included usage per seat, shared among org members
\nUsage past those included credits is billed on top of the subscription (pay-as-you-go)
\nOrganization admins can enable/disable usage of Inference Providers and set a spending limit (on top of included credits)
\nTeam & Enterprise orgs have a dedicated Inference Providers dashboard, offering full visibility into team usage across our serverless inference partners
\nMore info on pricing here: Pricing and Billing . We also have more info on the features of Team and Enterprise here: Hugging Face – Pricing.
\nHope this helps! Let me know if you have other questions.
', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T14:34:33.046Z', 'reply_count': 1, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/inference-providers/en/pricing', 'internal': False, 'reflection': False, 'title': 'Pricing and Billing', 'clicks': 1}, {'url': 'https://huggingface.co/changelog/inference-providers-dashboard', 'internal': False, 'reflection': False, 'title': 'New Inference Providers Dashboard', 'clicks': 0}, {'url': 'https://huggingface.co/pricing', 'internal': False, 'reflection': False, 'title': 'Hugging Face – Pricing', 'clicks': 0}, {'url': 'https://huggingface.co/enterprise', 'internal': False, 'reflection': False, 'title': 'Enterprise Hub - Hugging Face', 'clicks': 0}, {'url': 'https://huggingface.co/enterprise?subscribe=true', 'internal': False, 'reflection': False, 'title': 'Enterprise Hub - Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 101215, 'username': 'lrizzellotaskbase', 'name': 'Luca Rizzello', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/l/da6949/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/4', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237164, 'name': 'Luca Rizzello', 'username': 'lrizzellotaskbase', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/l/da6949/{size}.png', 'created_at': '2025-08-07T14:42:09.441Z', 'cooked': 'Thanks for the reply, but that still leaves my main question open: Is it possible to use huggingface’s pay-per-use inference (more specifically for Qwen Embedding 8B) as a company without having to upgrade to team or entreprise?
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T14:42:09.441Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'Luca Rizzello', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 31941, 'username': 'meganariley', 'name': 'Megan Riley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101215, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237172, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-08-07T15:03:10.956Z', 'cooked': 'A PRO, Team, or Enterprise subscription is needed - more here: Pricing and Billing .
', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T15:03:10.956Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/inference-providers/en/pricing#pay-as-you-go-details', 'internal': False, 'reflection': False, 'title': 'Pricing and Billing', 'clicks': 2}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 101215, 'username': 'lrizzellotaskbase', 'name': 'Luca Rizzello', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/l/da6949/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 237256, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-08T03:03:26.286Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-08-08T03:03:26.286Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 165927, 'topic_slug': 'can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/can-you-use-payg-for-an-entreprise-without-a-team-entreprise-plan/165927/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+I am currently trying to generate a lot of embeddings as part of a research project for my company.
+We have a team account setup as well as a valid billing method, and a token associated to our company in order to perform API calls.
+I’m using Qwen3-Embeddings-8B ( Qwen/Qwen3-Embedding-8B · Hugging Face )
+I can call it and get some embeddings, but after around 3000 or so embeddings I get hit with a limit and receive a 402 “Payment Required” exception. This surprised me since we do have a billing method.
+Then I looked into it a bit more and saw that “Inference Usage” has a max limit of $0 per month unless you have a team/entreprise account. So that means that you can’t pay per usage at all as a company until you set that up? Am I understading this correctly?
","A PRO, Team, or Enterprise subscription is needed - more here: Pricing and Billing .
" +Upload efficiently for lazy split download,https://discuss.huggingface.co/t/upload-efficiently-for-lazy-split-download/165834,165834,5,2025-08-06 10:06:02.849000+00:00,"[{'id': 236898, 'name': 'Élie Goudout', 'username': 'ego-thales', 'avatar_template': '/user_avatar/discuss.huggingface.co/ego-thales/{size}/52182_2.png', 'created_at': '2025-08-06T10:06:02.938Z', 'cooked': 'Hi everyone,
\nI’m a beginner regarding HuggigFace and I must say I’m completely lost in their tutorials.
\nEssentially CIFAR 10, structured as follows:
\ndata/airplane/airplane_xxxx.png\ndata/airplane/cat_yyyy.png\n...\nwhere xxxx goes from 0000 to 5999 and
0000 -> 0999 belong to test,1000 -> 5999 belong to train.To upload it with:
\nleave_out=""cat"" for example to treat cats separately).train, test and leftout.leave_out=""cat"", split=""leftout"", then HF only downloads the cat samples.I have trouble with the last part honestly…
\nI think from what I understood here that I need to create a custom dataset.py fils with the BuilderConfig and DatasetBuilder. But I have many questions:
\nclass Squad(datasets.GeneratorBasedBuilder):\n """"""SQUAD: The Stanford Question Answering Dataset. Version 1.1.""""""\n\n def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:\n downloaded_files = dl_manager.download_and_extract(_URLS)\n\n return [\n datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={""filepath"": downloaded_files[""train""]}),\n datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={""filepath"": downloaded_files[""dev""]}),\n ]\nseems to eagerly download every split??
\n2. I don’t really understand whether the script defining the DatasetBuilder will be used locally by me to upload to HF hub, or if it will be executed remotely by users and I should simply upload the raw files as I currently have tehm locally?
\n3. I think I can a maybe group files by test/train and class into zipballs to provide more efficient downloading? ut at this point it seems like I’m doing all the optimizing stuff HuggingFace should do for me?
Thanks in advance, it’s really hard to get into this from a beginner POV.
\nAl the best!
\nÉlie
\nI hav
Currently, your dataset has labels (such as “cat”) in the file names, but if you use directory (or archive file) names as labels instead of file names and organize them hierarchically, you should be able to organize the dataset hierarchically via ImageFolder.
\nIncidentally, ImageFolder does not seem to be very efficient when the dataset is huge.
\nhttps://github.com/huggingface/datasets/issues/5317
\n\n2
\n
I think the dataset builder script is executed locally.
\nBy the way, since executing the dataset builder directly from Hub is no longer recommended, it might be more convenient to publish the built data set if you want to make it public.
\n\n3
\n
Maybe true. I think it’s more convenient to divide them intentionally to a certain extent in some cases.
', 'post_number': 2, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-06T12:54:16.594Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 10.4, 'yours': False, 'topic_id': 165834, 'topic_slug': 'upload-efficiently-for-lazy-split-download', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/standard-way-to-upload-huge-dataset/81265', 'internal': True, 'reflection': False, 'title': 'Standard way to upload huge dataset', 'clicks': 0}, {'url': 'https://github.com/huggingface/datasets/issues/5317', 'internal': False, 'reflection': False, 'title': '`ImageFolder` performs poorly with large datasets · Issue #5317 · huggingface/datasets · GitHub', 'clicks': 0}, {'url': 'https://github.com/huggingface/datasets/issues/7693', 'internal': False, 'reflection': False, 'title': 'Dataset scripts are no longer supported, but found superb.py · Issue #7693 · huggingface/datasets · GitHub', 'clicks': 0}, {'url': 'https://github.com/huggingface/datasets/issues/5243', 'internal': False, 'reflection': False, 'title': 'Download only split data · Issue #5243 · huggingface/datasets · GitHub', 'clicks': 0}, {'url': 'https://huggingface.co/docs/datasets/en/image_dataset', 'internal': False, 'reflection': False, 'title': 'Create an image dataset', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/upload-efficiently-for-lazy-split-download/165834/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 236962, 'name': 'Élie Goudout', 'username': 'ego-thales', 'avatar_template': '/user_avatar/discuss.huggingface.co/ego-thales/{size}/52182_2.png', 'created_at': '2025-08-06T15:50:14.049Z', 'cooked': 'Thanks for your anwer and interesting pointers!
\nI am using ImageFolder structure currently but:
leave_out=""cat"" I mentioned)\n\nBy the way, since executing the dataset builder directly from Hub is no longer recommended,
\n
Hmmm that’s a bummer.
\n\n\nit might be more convenient to publish the built data set if you want to make it public.
\n
Could you explain what you mean by “built” please? Because when I browse other datasets, they never upload files like I did (it seems stupid to, so I expected that), they often use parquet (I don’t think it’s very appropriate for images? Maybe zip better?). Is that what you mean?
Or do you mean “built” as in “publish it 11 times with 11 strategies in 11 folders (entire dataset + 10 times minus one class)”?
\nAll the best.
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-06T15:51:17.519Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 165834, 'topic_slug': 'upload-efficiently-for-lazy-split-download', 'display_username': 'Élie Goudout', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/ego-thales/cifar10/tree/main', 'internal': False, 'reflection': False, 'title': 'ego-thales/cifar10 at main', 'clicks': 1}, {'url': 'https://github.com/huggingface/datasets/issues/7693', 'internal': False, 'reflection': False, 'title': 'Dataset scripts are no longer supported, but found superb.py · Issue #7693 · huggingface/datasets · GitHub', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101145, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/upload-efficiently-for-lazy-split-download/165834/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 237013, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-06T23:14:40.475Z', 'cooked': '\n\nI cannot get it to work with “calibration” split name
\n
In many cases, placing files and folders into the data folder works well.
\nFile names and splits
\n\nCould you explain what you mean by “built” please? Because when I browse other datasets, they never upload files like I did (it seems stupid to, so I expected that), they often use
\nparquet(I don’t think it’s very appropriate for images? Maybezipbetter?). Is that what you mean?
Yes. In parquet (default) or in WebDataset.
\n\nYes. In
\nparquet(default) or inWebDataset.
Ok thanks, I’ll eventually lean towards this.
\nRegarding the names, I know already that “calibration”, but following the tutorial for manual configuration with (metadata from my README.md)
configs:\n - config_name: default\n data_files:\n - split: train\n path: train/*/*.png\n - split: calibration\n path: calibration/*/*.png\n - split: test\n path: test/*/*.png\nI made it work now!
\nI think I’ll eventually settle for this, and use the filters option to leave_out specific classes on-the-fly. I cannot find the proper documentation for filters format though. I you have a pointer, that’d be lovely!
Again, thank you very much for your help!
\nAll the best.
\nI edited the original message as I made a typo in the manual config paths previously.
\nSecond edit, I still had a typo, now it seems to work!
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-08-07T09:09:12.824Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 20.4, 'yours': False, 'topic_id': 165834, 'topic_slug': 'upload-efficiently-for-lazy-split-download', 'display_username': 'Élie Goudout', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/datasets/v4.0.0/en/image_load#webdataset', 'internal': False, 'reflection': False, 'title': 'Load image data', 'clicks': 0}, {'url': 'https://huggingface.co/docs/hub/en/datasets-manual-configuration', 'internal': False, 'reflection': False, 'title': 'Manual Configuration', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 101145, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/upload-efficiently-for-lazy-split-download/165834/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 237115, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-08-07T10:10:34.230Z', 'cooked': 'Great!
Since many people use .filter, I don’t know much about filters option, but it seems that they need to be passed in PyArrow format.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-08-07T22:11:20.225Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 165834, 'topic_slug': 'upload-efficiently-for-lazy-split-download', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/upload-efficiently-for-lazy-split-download/165834/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I’m a beginner regarding HuggigFace and I must say I’m completely lost in their tutorials.
+Essentially CIFAR 10, structured as follows:
+data/airplane/airplane_xxxx.png
+data/airplane/cat_yyyy.png
+...
+where xxxx goes from 0000 to 5999 and
0000 -> 0999 belong to test,1000 -> 5999 belong to train.To upload it with:
+leave_out=""cat"" for example to treat cats separately).train, test and leftout.leave_out=""cat"", split=""leftout"", then HF only downloads the cat samples.I have trouble with the last part honestly…
+I think from what I understood here that I need to create a custom dataset.py fils with the BuilderConfig and DatasetBuilder. But I have many questions:
+class Squad(datasets.GeneratorBasedBuilder):
+ """"""SQUAD: The Stanford Question Answering Dataset. Version 1.1.""""""
+
+ def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
+ downloaded_files = dl_manager.download_and_extract(_URLS)
+
+ return [
+ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={""filepath"": downloaded_files[""train""]}),
+ datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={""filepath"": downloaded_files[""dev""]}),
+ ]
+seems to eagerly download every split??
+2. I don’t really understand whether the script defining the DatasetBuilder will be used locally by me to upload to HF hub, or if it will be executed remotely by users and I should simply upload the raw files as I currently have tehm locally?
+3. I think I can a maybe group files by test/train and class into zipballs to provide more efficient downloading? ut at this point it seems like I’m doing all the optimizing stuff HuggingFace should do for me?
Thanks in advance, it’s really hard to get into this from a beginner POV.
+Al the best!
+Élie
+I hav
++Yes. In
+parquet(default) or inWebDataset.
Ok thanks, I’ll eventually lean towards this.
+Regarding the names, I know already that “calibration”, but following the tutorial for manual configuration with (metadata from my README.md)
configs:
+ - config_name: default
+ data_files:
+ - split: train
+ path: train/*/*.png
+ - split: calibration
+ path: calibration/*/*.png
+ - split: test
+ path: test/*/*.png
+I made it work now!
+I think I’ll eventually settle for this, and use the filters option to leave_out specific classes on-the-fly. I cannot find the proper documentation for filters format though. I you have a pointer, that’d be lovely!
Again, thank you very much for your help!
+All the best.
+I edited the original message as I made a typo in the manual config paths previously.
+Second edit, I still had a typo, now it seems to work!
" +The effect of padding_side,https://discuss.huggingface.co/t/the-effect-of-padding-side/67188,67188,9,2023-12-27 16:32:44.724000+00:00,"[{'id': 105773, 'name': 'zhouzaida', 'username': 'zhouzaida', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/ce7236/{size}.png', 'created_at': '2023-12-27T16:32:44.782Z', 'cooked': 'Hello, I have a question about the documentation here (Generation with LLMs). Below is a code block, and I’m curious why setting padding_side to ‘left’ yields the correct inference result, while setting it to ‘right’ does not work. The attention_mask is also passed to the model’s generate method, so theoretically, it should be able to correctly infer the next token.
# The tokenizer initialized above has right-padding active by default: the 1st sequence,\n# which is shorter, has padding on the right side. Generation fails to capture the logic.\nmodel_inputs = tokenizer(\n [""1, 2, 3"", ""A, B, C, D, E""], padding=True, return_tensors=""pt""\n).to(""cuda"")\ngenerated_ids = model.generate(**model_inputs)\ntokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n\n# With left-padding, it works as expected!\ntokenizer = AutoTokenizer.from_pretrained(""mistralai/Mistral-7B-v0.1"", padding_side=""left"")\ntokenizer.pad_token = tokenizer.eos_token # Most LLMs don\'t have a pad token by default\nmodel_inputs = tokenizer(\n [""1, 2, 3"", ""A, B, C, D, E""], padding=True, return_tensors=""pt""\n).to(""cuda"")\ngenerated_ids = model.generate(**model_inputs)\ntokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\nHi,
\nThis is explained here: Generation with LLMs.
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 15, 'updated_at': '2023-12-27T19:57:53.146Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 118, 'reads': 453, 'readers_count': 452, 'score': 730.2, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/llm_tutorial#wrong-padding-side', 'internal': False, 'reflection': False, 'title': 'Generation with LLMs', 'clicks': 1603}, {'url': 'https://huggingface.co/learn/nlp-course/chapter1/6?fw=pt', 'internal': False, 'reflection': False, 'title': 'Decoder models - Hugging Face NLP Course', 'clicks': 93}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 3}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 105841, 'name': 'zhouzaida', 'username': 'zhouzaida', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/ce7236/{size}.png', 'created_at': '2023-12-28T02:14:27.175Z', 'cooked': 'LLMs are decoder-only architectures, meaning they continue to iterate on your input prompt. If your inputs do not have the same length, they need to be padded. Since LLMs are not trained to continue from pad tokens, your input needs to be left-padded.
\n
Hi @nielsr , thanks for your reply. I understand the role of padding, the point that actually confused me was why padding right affects the output of the model, since the attention mask has already been passed in, the padding should be masked out in atten_weight, and theoretically it shouldn’t have an effect.
', 'post_number': 3, 'post_type': 1, 'posts_count': 15, 'updated_at': '2023-12-28T02:14:27.175Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 64, 'reads': 426, 'readers_count': 425, 'score': 419.8, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'zhouzaida', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 36936, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 105860, 'name': 'zhouzaida', 'username': 'zhouzaida', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/ce7236/{size}.png', 'created_at': '2023-12-28T06:30:38.786Z', 'cooked': '@nielsr thanks for your help. After debugging the code, I found the key to the unexpected behavior (padding_side=‘right’) is the next_token comeing from the logit of pad token. I thought it would somehow get the logit of the last non-pad token as the predicted next token, but that’s not actually the case, it simply takes the last token (which could be a pad token).
\n while True:\n if synced_gpus:\n # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.\n # The following logic allows an early break if all peers finished generating their sequence\n this_peer_finished_flag = torch.tensor(0.0 if this_peer_finished else 1.0).to(input_ids.device)\n # send 0.0 if we finished, 1.0 otherwise\n dist.all_reduce(this_peer_finished_flag, op=dist.ReduceOp.SUM)\n # did all peers finish? the reduced sum will be 0.0 then\n if this_peer_finished_flag.item() == 0.0:\n break\n\n # prepare model inputs\n model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)\n\n # forward pass to get next token\n outputs = self(\n **model_inputs,\n return_dict=True,\n output_attentions=output_attentions,\n output_hidden_states=output_hidden_states,\n )\n\n if synced_gpus and this_peer_finished:\n continue # don\'t waste resources running the code we don\'t need\n\n next_token_logits = outputs.logits[:, -1, :]\nHi dude, I couldnt quite understand the logic here
\nAnd one more thing: I saw this piece of code:
\n\ndecided to pad on left side but with eos token ? Don’t the models automatically stop when they see eos tokens? Shouldn’t there be a problem here?
', 'post_number': 5, 'post_type': 1, 'posts_count': 15, 'updated_at': '2024-05-17T23:56:44.806Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 111, 'reads': 270, 'readers_count': 269, 'score': 628.6, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Doğan Keskin', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://us1.discourse-cdn.com/hellohellohello/original/3X/7/f/7f5e5874b3428578ac8c05c7572d269444bbde4b.png', 'internal': False, 'reflection': False, 'title': '7f5e5874b3428578ac8c05c7572d269444bbde4b.png', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 36936, 'username': 'zhouzaida', 'name': 'zhouzaida', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/ce7236/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 50459, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/5', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 131907, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2024-05-20T09:42:01.346Z', 'cooked': 'Hi,
\nIf models don’t have a padding token set one can use the EOS token as padding token, and pad from the left at inference time.
\nThis is not an issue since the model will then see “<eos> <eos> <eos> (…) hello your name is” => then the model is prompted to continue the token “is”, so it will generate several new tokens until it will generate an EOS token.
', 'post_number': 6, 'post_type': 1, 'posts_count': 15, 'updated_at': '2024-05-21T07:00:32.905Z', 'reply_count': 1, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 38, 'reads': 232, 'readers_count': 231, 'score': 281.0, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 50459, 'username': 'DoganK01', 'name': 'Doğan Keskin', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/5fc32e/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/6', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 3}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 131984, 'name': 'Doğan Keskin', 'username': 'DoganK01', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/5fc32e/{size}.png', 'created_at': '2024-05-20T21:39:48.181Z', 'cooked': 'is it like [EOS, EOS, EOS, Hello, your, name, is, … ]? Because in this format, model should stop since it sees the stop token. what is I’m missing ?
', 'post_number': 7, 'post_type': 1, 'posts_count': 15, 'updated_at': '2024-05-20T21:39:48.181Z', 'reply_count': 1, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 22, 'reads': 218, 'readers_count': 217, 'score': 173.2, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Doğan Keskin', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 50459, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/7', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 132060, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2024-05-21T07:00:58.277Z', 'cooked': 'Yes, sorry for Forum was hiding the <eos> tokens in my reply
I didnt understand, what is the specific reason to use EOS to do padding it? Why we using EOS? and why left side? isn’t it the case that model stops when it sees the EOS token generated from itsel? (for example [BOS] Hi, how are you? [EOS]). For this example, shouldnt the model just stop since the model generated [EOS] token when the model tokenized “?” ?
\nIt makes sense to use the EOS token when we set the padding side = right. Likewise, we can also use BOS (begin of sentece) tokens for padding, right? And it makes sense when we set the padding side = left. What am I missing?
', 'post_number': 9, 'post_type': 1, 'posts_count': 15, 'updated_at': '2024-05-21T23:37:19.990Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 33, 'reads': 203, 'readers_count': 202, 'score': 230.2, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Doğan Keskin', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 50459, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 137937, 'name': 'Kalpan Mukherjee', 'username': 'kalpanmukherjee', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/k/90ced4/{size}.png', 'created_at': '2024-06-15T18:23:52.303Z', 'cooked': '@DoganK01 from what I understand what happens is the model sees -
\n[eos] - nothing to generate
\n[eos] [eos] - nothing to generate
\n[eos] [eos] hello - generates logits for after hello
hope this clears it up for you!
', 'post_number': 10, 'post_type': 1, 'posts_count': 15, 'updated_at': '2024-06-15T18:23:52.303Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 32, 'reads': 168, 'readers_count': 167, 'score': 208.6, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Kalpan Mukherjee', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 54252, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/10', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 154686, 'name': 'Weikang Qiu', 'username': 'Boltzmachine', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/46a35a/{size}.png', 'created_at': '2024-09-10T16:52:45.385Z', 'cooked': 'I cannot understand why huggingface implement like this. Why don’t they extract the last non-pad tokens of each sample?
', 'post_number': 11, 'post_type': 1, 'posts_count': 15, 'updated_at': '2024-09-10T16:52:45.385Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 27, 'reads': 115, 'readers_count': 114, 'score': 168.0, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Weikang Qiu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 36936, 'username': 'zhouzaida', 'name': 'zhouzaida', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/ce7236/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 1864, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/11', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 193934, 'name': 'Robin Lee', 'username': 'rlee002', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/85f322/{size}.png', 'created_at': '2025-01-07T02:45:52.618Z', 'cooked': 'Adding onto here, I believe this is only for the generation side (inference side) of the model. So for fine-tuning an LLM, do we still keep the right padding or do we follow the same logic as for inference and keep the left padding?
', 'post_number': 12, 'post_type': 1, 'posts_count': 15, 'updated_at': '2025-01-07T02:45:52.618Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 65, 'readers_count': 64, 'score': 148.0, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Robin Lee', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 24692, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/12', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}, {'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216378, 'name': 'Mauro Camara Escudero', 'username': 'MauroExtrac', 'avatar_template': '/user_avatar/discuss.huggingface.co/mauroextrac/{size}/38514_2.png', 'created_at': '2025-04-17T15:55:22.888Z', 'cooked': 'Did you ever find out?
', 'post_number': 13, 'post_type': 1, 'posts_count': 15, 'updated_at': '2025-04-17T15:55:22.888Z', 'reply_count': 0, 'reply_to_post_number': 12, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 34, 'readers_count': 33, 'score': 61.8, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Mauro Camara Escudero', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 24692, 'username': 'rlee002', 'name': 'Robin Lee', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/85f322/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 78649, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/13', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 224304, 'name': 'Doğan Keskin', 'username': 'DoganK01', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/5fc32e/{size}.png', 'created_at': '2025-05-27T12:35:17.860Z', 'cooked': 'Guys, I figured it out. Since models are decoder-only (autoregressive), its nonsense applying padding on right side. Because model predicts the next token by looking at last as you can figure this out @zhouzaida s last answer in this thread. And about model stopping predicting next token when it sees EOS, its just adjusting it in the code by telling model that it shouldnt focus on padding (EOS) tokens in the beginning and then should skip them. This is what I’ve figured out. But when we say model to skip those padding tokens, it shouldnt have any importance to set pad token to EOS or BOS. I dont have answer for the last one
This is indeed the root cause. IMO this can be easily fixed (i.e., by taking the logits of the last non-padding token); not sure why it’s not implemented this way in the first place.
', 'post_number': 15, 'post_type': 1, 'posts_count': 15, 'updated_at': '2025-08-07T16:21:19.415Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 9, 'readers_count': 8, 'score': 26.8, 'yours': False, 'topic_id': 67188, 'topic_slug': 'the-effect-of-padding-side', 'display_username': 'Jingyang Zhang', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 36936, 'username': 'zhouzaida', 'name': 'zhouzaida', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/ce7236/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 30869, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-effect-of-padding-side/67188/15', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello, I have a question about the documentation here (Generation with LLMs). Below is a code block, and I’m curious why setting padding_side to ‘left’ yields the correct inference result, while setting it to ‘right’ does not work. The attention_mask is also passed to the model’s generate method, so theoretically, it should be able to correctly infer the next token.
# The tokenizer initialized above has right-padding active by default: the 1st sequence,
+# which is shorter, has padding on the right side. Generation fails to capture the logic.
+model_inputs = tokenizer(
+ [""1, 2, 3"", ""A, B, C, D, E""], padding=True, return_tensors=""pt""
+).to(""cuda"")
+generated_ids = model.generate(**model_inputs)
+tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+
+# With left-padding, it works as expected!
+tokenizer = AutoTokenizer.from_pretrained(""mistralai/Mistral-7B-v0.1"", padding_side=""left"")
+tokenizer.pad_token = tokenizer.eos_token # Most LLMs don't have a pad token by default
+model_inputs = tokenizer(
+ [""1, 2, 3"", ""A, B, C, D, E""], padding=True, return_tensors=""pt""
+).to(""cuda"")
+generated_ids = model.generate(**model_inputs)
+tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+@nielsr thanks for your help. After debugging the code, I found the key to the unexpected behavior (padding_side=‘right’) is the next_token comeing from the logit of pad token. I thought it would somehow get the logit of the last non-pad token as the predicted next token, but that’s not actually the case, it simply takes the last token (which could be a pad token).
+ while True:
+ if synced_gpus:
+ # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
+ # The following logic allows an early break if all peers finished generating their sequence
+ this_peer_finished_flag = torch.tensor(0.0 if this_peer_finished else 1.0).to(input_ids.device)
+ # send 0.0 if we finished, 1.0 otherwise
+ dist.all_reduce(this_peer_finished_flag, op=dist.ReduceOp.SUM)
+ # did all peers finish? the reduced sum will be 0.0 then
+ if this_peer_finished_flag.item() == 0.0:
+ break
+
+ # prepare model inputs
+ model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)
+
+ # forward pass to get next token
+ outputs = self(
+ **model_inputs,
+ return_dict=True,
+ output_attentions=output_attentions,
+ output_hidden_states=output_hidden_states,
+ )
+
+ if synced_gpus and this_peer_finished:
+ continue # don't waste resources running the code we don't need
+
+ next_token_logits = outputs.logits[:, -1, :]
+I’m using AutoTrain for training my models, I’m currently training llama_3.1_8B with my data but I have always trained different models when I added new data on my dataset, so I basically have re-trained another llama_3.1_8B and I thought this is not the best practice…
\nSo I decided to re-train the same model I have trained before with my data and I thought that on the form where I put the model I want to train, I should point to my model hf repo and when I start the training the status is success, but right when the training effectively starts it raises this error:
ValueError: Unrecognized model in DigioMatthy/the-name-of-my-model Should have a `model_type` key in its config.json, or contain one of the following strings in its name: albert, align, altclip, aria, aria_text, audio-spectrogram-transformer, autoformer,\nbamba, bark, bart, beit, bert, bert-generation, big_bird, bigbird_pegasus, biogpt, bit, blenderbot, blenderbot-small, blip, blip-2, bloom, bridgetower, bros, camembert, canine, chameleon, chinese_clip, chinese_clip_vision_model, clap, clip,\nclip_text_model, clip_vision_model, clipseg, clvp, code_llama, codegen, cohere, cohere2, colpali, conditional_detr, convbert, convnext, convnextv2, cpmant, ctrl, cvt, dac, data2vec-audio, data2vec-text, data2vec-vision, dbrx, deberta, deberta-v2,\ndecision_transformer, deformable_detr, deit, depth_anything, deta, detr, diffllama, dinat, dinov2, dinov2_with_registers, distilbert, donut-swin, dpr, dpt, efficientformer, efficientnet, electra, emu3, encodec, encoder-decoder, ernie, ernie_m, esm,\nfalcon, falcon_mamba, fastspeech2_conformer, flaubert, flava, fnet, focalnet, fsmt, funnel, fuyu, gemma, gemma2, git, glm, glpn, gpt-sw3, gpt2, gpt_bigcode, gpt_neo, gpt_neox, gpt_neox_japanese, gptj, gptsan-japanese, granite, granitemoe, graphormer,\ngrounding-dino, groupvit, hiera, hubert, ibert, idefics, idefics2, idefics3, idefics3_vision, ijepa, imagegpt, informer, instructblip, instructblipvideo, jamba, jetmoe, jukebox, kosmos-2, layoutlm, layoutlmv2, layoutlmv3, led, levit, lilt, llama, llava,\nllava_next, llava_next_video, llava_onevision, longformer, longt5, luke, lxmert, m2m_100, mamba, mamba2, marian, markuplm, mask2former, maskformer, maskformer-swin, mbart, mctct, mega, megatron-bert, mgp-str, mimi, mistral, mixtral, mllama, mobilebert,\nmobilenet_v1, mobilenet_v2, mobilevit, mobilevitv2, modernbert, moonshine, moshi, mpnet, mpt, mra, mt5, musicgen, musicgen_melody, mvp, nat, nemotron, nezha, nllb-moe, nougat, nystromformer, olmo, olmo2, olmoe, omdet-turbo, oneformer, open-llama, openai-gpt,\nopt, owlv2, owlvit, paligemma, patchtsmixer, patchtst, pegasus, pegasus_x, perceiver, persimmon, phi, phi3, phimoe, pix2struct, pixtral, plbart, poolformer, pop2piano, prophetnet, pvt, pvt_v2, qdqbert, qwen2, qwen2_audio, qwen2_audio_encoder,\nqwen2_moe, qwen2_vl, rag, realm, recurrent_gemma, reformer, regnet, rembert, resnet, retribert, roberta, roberta-prelayernorm, roc_bert, roformer, rt_detr, rt_detr_resnet, rwkv, sam, seamless_m4t, seamless_m4t_v2, segformer, seggpt, sew, sew-d, siglip,\nsiglip_vision_model, speech-encoder-decoder, speech_to_text, speech_to_text_2, speecht5, splinter, squeezebert, stablelm, starcoder2, superpoint, swiftformer, swin, swin2sr, swinv2, switch_transformers, t5, table-transformer, tapas, textnet,\ntime_series_transformer, timesformer, timm_backbone, timm_wrapper, trajectory_transformer, transfo-xl, trocr, tvlt, tvp, udop, umt5, unispeech, unispeech-sat, univnet, upernet, van, video_llava, videomae, vilt, vipllava, vision-encoder-decoder, vision-text-dual-encoder,\nvisual_bert, vit, vit_hybrid, vit_mae, vit_msn, vitdet, vitmatte, vitpose, vitpose_backbone, vits, vivit, wav2vec2, wav2vec2-bert, wav2vec2-conformer, wavlm, whisper, xclip, xglm, xlm, xlm-prophetnet, xlm-roberta,\nxlm-roberta-xl, xlnet, xmod, yolos, yoso, zamba, zoedepth\nAm I missing something?
\nIt has to be a way to re-train the same model (with AutoTrain) on new data without forgetting.
The immediate cause is that config.json cannot be found. There are several possible reasons for this, but if the repository was created with AutoTrainAdvanced, it may be because only the adapter is saved instead of the entire model.
Yes I can confirm that what is gonna save after the training are just adapters, infact I have written a script that merge these adapters with the original model’s weights and after that I can convert it to .gguf in order to upload it on Ollama.
\nI imagined that this ValueError was due to this fact.
\nIn your opinion, should I use the same script as I said before, but just adding at the end of the code something that it will push the entire model merged on my hf hub?
Yeah. If it can be converted to GGUF, I think save_pretrained has probably been completed, so you should be able to use it as a fine-tuning model just by uploading it.
If you want to save the complete model instead of the adapter for future training, you should be able to do so by just specifying --merge_adapter.
Oh wait, do you mean that on AutoTrain I can set merge adapter?
\nHow can I do it?
\nI just have these parameters in this way (if I enable JSON):
\n
I think you just need to set ""merge_adapter"": ""true""… Probably.
OMG! Yes it works!!! Thank you so much!!!
\nJust seeing that if I directly save the entire model after the training with ""merge_adapter"": ""true"" and I explore the model files inside the repo, the safetensors are 4, while when I was merging the model manually with the script the safetensors were 7, it’s not a problem, because right when you download the model with a script that just takes the model and tokenizer from a repo that contains the entire model, it will have all the safetensors!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-08-06T00:16:29.369Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 165704, 'topic_slug': 'how-can-i-update-knowledge-of-a-model-already-trained-before-valueerror-unrecognized-model', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-can-i-update-knowledge-of-a-model-already-trained-before-valueerror-unrecognized-model/165704/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m using AutoTrain for training my models, I’m currently training llama_3.1_8B with my data but I have always trained different models when I added new data on my dataset, so I basically have re-trained another llama_3.1_8B and I thought this is not the best practice…
+So I decided to re-train the same model I have trained before with my data and I thought that on the form where I put the model I want to train, I should point to my model hf repo and when I start the training the status is success, but right when the training effectively starts it raises this error:
ValueError: Unrecognized model in DigioMatthy/the-name-of-my-model Should have a `model_type` key in its config.json, or contain one of the following strings in its name: albert, align, altclip, aria, aria_text, audio-spectrogram-transformer, autoformer,
+bamba, bark, bart, beit, bert, bert-generation, big_bird, bigbird_pegasus, biogpt, bit, blenderbot, blenderbot-small, blip, blip-2, bloom, bridgetower, bros, camembert, canine, chameleon, chinese_clip, chinese_clip_vision_model, clap, clip,
+clip_text_model, clip_vision_model, clipseg, clvp, code_llama, codegen, cohere, cohere2, colpali, conditional_detr, convbert, convnext, convnextv2, cpmant, ctrl, cvt, dac, data2vec-audio, data2vec-text, data2vec-vision, dbrx, deberta, deberta-v2,
+decision_transformer, deformable_detr, deit, depth_anything, deta, detr, diffllama, dinat, dinov2, dinov2_with_registers, distilbert, donut-swin, dpr, dpt, efficientformer, efficientnet, electra, emu3, encodec, encoder-decoder, ernie, ernie_m, esm,
+falcon, falcon_mamba, fastspeech2_conformer, flaubert, flava, fnet, focalnet, fsmt, funnel, fuyu, gemma, gemma2, git, glm, glpn, gpt-sw3, gpt2, gpt_bigcode, gpt_neo, gpt_neox, gpt_neox_japanese, gptj, gptsan-japanese, granite, granitemoe, graphormer,
+grounding-dino, groupvit, hiera, hubert, ibert, idefics, idefics2, idefics3, idefics3_vision, ijepa, imagegpt, informer, instructblip, instructblipvideo, jamba, jetmoe, jukebox, kosmos-2, layoutlm, layoutlmv2, layoutlmv3, led, levit, lilt, llama, llava,
+llava_next, llava_next_video, llava_onevision, longformer, longt5, luke, lxmert, m2m_100, mamba, mamba2, marian, markuplm, mask2former, maskformer, maskformer-swin, mbart, mctct, mega, megatron-bert, mgp-str, mimi, mistral, mixtral, mllama, mobilebert,
+mobilenet_v1, mobilenet_v2, mobilevit, mobilevitv2, modernbert, moonshine, moshi, mpnet, mpt, mra, mt5, musicgen, musicgen_melody, mvp, nat, nemotron, nezha, nllb-moe, nougat, nystromformer, olmo, olmo2, olmoe, omdet-turbo, oneformer, open-llama, openai-gpt,
+opt, owlv2, owlvit, paligemma, patchtsmixer, patchtst, pegasus, pegasus_x, perceiver, persimmon, phi, phi3, phimoe, pix2struct, pixtral, plbart, poolformer, pop2piano, prophetnet, pvt, pvt_v2, qdqbert, qwen2, qwen2_audio, qwen2_audio_encoder,
+qwen2_moe, qwen2_vl, rag, realm, recurrent_gemma, reformer, regnet, rembert, resnet, retribert, roberta, roberta-prelayernorm, roc_bert, roformer, rt_detr, rt_detr_resnet, rwkv, sam, seamless_m4t, seamless_m4t_v2, segformer, seggpt, sew, sew-d, siglip,
+siglip_vision_model, speech-encoder-decoder, speech_to_text, speech_to_text_2, speecht5, splinter, squeezebert, stablelm, starcoder2, superpoint, swiftformer, swin, swin2sr, swinv2, switch_transformers, t5, table-transformer, tapas, textnet,
+time_series_transformer, timesformer, timm_backbone, timm_wrapper, trajectory_transformer, transfo-xl, trocr, tvlt, tvp, udop, umt5, unispeech, unispeech-sat, univnet, upernet, van, video_llava, videomae, vilt, vipllava, vision-encoder-decoder, vision-text-dual-encoder,
+visual_bert, vit, vit_hybrid, vit_mae, vit_msn, vitdet, vitmatte, vitpose, vitpose_backbone, vits, vivit, wav2vec2, wav2vec2-bert, wav2vec2-conformer, wavlm, whisper, xclip, xglm, xlm, xlm-prophetnet, xlm-roberta,
+xlm-roberta-xl, xlnet, xmod, yolos, yoso, zamba, zoedepth
+Am I missing something?
+It has to be a way to re-train the same model (with AutoTrain) on new data without forgetting.
I think you just need to set ""merge_adapter"": ""true""… Probably.
I’m unable to download gated models (e.g., mistralai/Mistral-7B-Instruct-v0.2) using huggingface_hub from within a Databricks cluster. Despite setting HF_HUB_DISABLE_XET=1 and removing any hf-xet or hf_transfer packages, the library continues attempting to contact cas-bridge.xethub.hf.co, which results in a repeated “RuntimeError: Data processing error: CAS service error : ReqwestMiddleware Error: Request failed after 5 retries”
HF_HUB_DISABLE_XET, HF_HUB_ENABLE_HF_TRANSFER)huggingface_hub to versions like 0.21.4, 0.23.0, and 0.30.2hf-xet is not installed (pip list, !find ~/.cache -name \'xet\')hf_hub_download as well — same issuehf-xet to latest version - still the same errorIt is unclear whether the cause is the same, but similar errors seem to have been reported.
', 'post_number': 2, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-07-28T10:09:22.277Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 13, 'readers_count': 12, 'score': 7.6, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/xet-core/issues/407#issuecomment-3117966733', 'internal': False, 'reflection': False, 'title': 'Cannot download file from XET hosted repo using CLI · Issue #407 · huggingface/xet-core · GitHub', 'clicks': 30}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235331, 'name': 'Manjunatha B', 'username': 'manjusavanth', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/c4cdca/{size}.png', 'created_at': '2025-07-28T13:06:17.184Z', 'cooked': 'that is correct, it is exactly the same error reported by GohioAC here
', 'post_number': 3, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-07-28T13:06:17.184Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 12, 'readers_count': 11, 'score': 22.4, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Manjunatha B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/xet-core/issues/407#issuecomment-3117966733', 'internal': False, 'reflection': False, 'title': 'Cannot download file from XET hosted repo using CLI · Issue #407 · huggingface/xet-core · GitHub', 'clicks': 14}, {'url': 'https://github.com/GohioAC', 'internal': False, 'reflection': False, 'title': 'GohioAC (Aritra Chatterjee) · GitHub', 'clicks': 9}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100468, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235433, 'name': 'Jared Sulzdorf', 'username': 'jsulz', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png', 'created_at': '2025-07-29T03:13:04.711Z', 'cooked': 'Hi @manjusavanth thanks for the report - Xet team member here.
\nThis does seem related to a few issues we’ve encountered recently, although you should be able to fall back to HTTP download through HF_HUB_DISABLE_XET=1.
How are you downloading mistralai/Mistral-7B-Instruct-v0.2? Is it through the huggingface-cli or one of the core Python function (e.g., snapshot_download)?
Could you tell me anything more about the Databricks environment?
', 'post_number': 4, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-07-29T03:13:04.711Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 12, 'readers_count': 11, 'score': 32.4, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Jared Sulzdorf', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 54269, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/4', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235440, 'name': 'Manjunatha B', 'username': 'manjusavanth', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/c4cdca/{size}.png', 'created_at': '2025-07-29T04:25:11.924Z', 'cooked': 'Hi @jsulz I have tried using HF_HUB_DISABLE_XET=1, this does not work for me.
\nBelow is the complete code:
\n%pip uninstall -y hf-xet huggingface_hub
\n%pip install huggingface-hub
\n%pip install hf_xet==v1.1.6rc2
\n%pip install vllm==0.8.5
\nimport os
\nfrom huggingface_hub import login
\nlogin(token=“token_id”)
from vllm import *
\n! python -m vllm.entrypoints.openai.api_server --model mistralai/Magistral-Small-2506 --dtype float16 --tensor-parallel-size 4 --port 8003 --max_model_len 15000 --tokenizer-mode “mistral”
on Databricks, I have run the code on clusters of V100 and T4 GPUs. These are the cluster spinned dedicatedly for the ML job without having pre-installed python packages.
', 'post_number': 5, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-07-29T04:25:11.924Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 12, 'readers_count': 11, 'score': 52.4, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Manjunatha B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 54269, 'username': 'jsulz', 'name': 'Jared Sulzdorf', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100468, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235595, 'name': 'Jared Sulzdorf', 'username': 'jsulz', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png', 'created_at': '2025-07-29T21:38:11.785Z', 'cooked': 'Thanks for those details @manjusavanth
\nBased on what I see here, you uninstall hf-xet but then reinstall it on line three (%pip install hf_xet==v1.1.6rc2). Regardless, the HF_HUB_DISABLE_XET flag, when turned on, should work. The issue with the flag may be related to this issue on the huggingface_hub repo. I would suggest posting about your experiences there as well.
As for the runtime error you are encountering, I believe that is related to a known issue we are seeing with the vllm library. You should be able to get around that by falling back to HTTP download with HF_HUB_DISABLE_XET (which appears to not work for you at the moment) or uninstalling hf-xet. If the HF_HUB_DISABLE_XET flag is not working for you, I would suggest running pip uninstall -y hf-xet after the installation of huggingface-hub and do not reinstall it.
I’ll follow up here once the hf-xet issue with vllm is addressed, and let me know if you have any questions.
@manjusavanth we believe we’ve addressed the root cause of the CAS service error you were seeing. You can pip install a release candidate for testing. I.e.,
pip install hf-xet==1.1.6rc5
Hi @jsulz , I have tried with pip install hf-xet==1.1.6rc5, this gives the same error as earlier. I changed nothing else apart from this line pip install hf-xet==1.1.6rc5.
', 'post_number': 8, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-07-30T06:12:35.574Z', 'reply_count': 1, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 21.8, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Manjunatha B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 54269, 'username': 'jsulz', 'name': 'Jared Sulzdorf', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100468, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/8', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235697, 'name': 'Jared Sulzdorf', 'username': 'jsulz', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png', 'created_at': '2025-07-30T14:33:53.687Z', 'cooked': 'Thanks for testing @manjusavanth! We’ll keep investigating.
\nTo make sure you’re unblocked and can download mistralai/Mistral-7B-Instruct-v0.2 did you see my earlier comment with respect to how you are loading in hf-xet?
I would review your code to ensure that either hf-xet is not installed and/or your environment recognizes the HF_HUB_DISABLE_XET. If, for whatever reason, HF_HUB_DISABLE_XET isn’t working for you, I would add your reproduction steps to the GitHub issue.
Hi @jsulz I did try to install huggingface-hub first and then uninstalling the hf-xet. Also set the flag “HF_HUB_DISABLE_XET” to 1. But I continue to receive the same error.
\nI also check for the presence of xet after uninstaaling, there is no xet, but the CAS error continues.
\nimport os
\nimport glob
\nxet_bin = glob.glob(os.path.expanduser(“~/.cache/huggingface/hub/extensions/**/xet”), recursive=True)
\nprint(“XET binaries found:”, xet_bin)
XET binaries found:
', 'post_number': 10, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-07-31T11:21:59.780Z', 'reply_count': 1, 'reply_to_post_number': 9, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 9, 'readers_count': 8, 'score': 31.8, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Manjunatha B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 54269, 'username': 'jsulz', 'name': 'Jared Sulzdorf', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100468, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': 'checklist change', 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/10', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235998, 'name': 'Jared Sulzdorf', 'username': 'jsulz', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png', 'created_at': '2025-08-01T16:01:01.887Z', 'cooked': 'I believe the issue with HF_HUB_DISABLE_XET may be related to the issue here HF_HUB_DISABLE_XET not disabling XET-based downloads · Issue #3266 · huggingface/huggingface_hub · GitHub
Can you confirm that you set the environment variable before you load the huggingface_hub library?
hi @jsulz I have tried setting the flag for HF_HUB_DISABLE_XET both before and after importing the huggingface_hub library, nothing seems to change as I get the same CAS error, this issue has become a pain as I have not been able to download the model for last 20days. I am not sure vLLM is adding to the issue.
This turned out to be the ip whitelisting issue. After getting the below urls whitelisted, the model download worked with xet.
\n', 'post_number': 13, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-08-04T16:11:00.097Z', 'reply_count': 1, 'reply_to_post_number': 12, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 7, 'readers_count': 6, 'score': 101.4, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Manjunatha B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://transfer.xethub.hf.co', 'internal': False, 'reflection': False, 'clicks': 35}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 100468, 'username': 'manjusavanth', 'name': 'Manjunatha B', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/c4cdca/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100468, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/13', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 236536, 'name': 'Jared Sulzdorf', 'username': 'jsulz', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png', 'created_at': '2025-08-04T17:10:39.538Z', 'cooked': '@manjusavanth ah, I’m sorry, that should’ve been the first thing I asked
Glad you resolved this and sorry for the runaround.
', 'post_number': 14, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-08-04T17:10:39.538Z', 'reply_count': 1, 'reply_to_post_number': 13, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 21.4, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Jared Sulzdorf', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 100468, 'username': 'manjusavanth', 'name': 'Manjunatha B', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/c4cdca/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 54269, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/14', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 236612, 'name': 'Manjunatha B', 'username': 'manjusavanth', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/c4cdca/{size}.png', 'created_at': '2025-08-05T06:36:59.483Z', 'cooked': 'Thank you for your time and guidance.
', 'post_number': 15, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-08-05T06:36:59.483Z', 'reply_count': 0, 'reply_to_post_number': 14, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 21.6, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'Manjunatha B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 54269, 'username': 'jsulz', 'name': 'Jared Sulzdorf', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100468, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/15', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 236801, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-05T18:37:34.342Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 16, 'post_type': 3, 'posts_count': 16, 'updated_at': '2025-08-05T18:37:34.342Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 164793, 'topic_slug': 'cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cas-service-error-when-downloading-gated-models-on-databricks-even-with-hf-hub-disable-xet-1/164793/16', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m unable to download gated models (e.g., mistralai/Mistral-7B-Instruct-v0.2) using huggingface_hub from within a Databricks cluster. Despite setting HF_HUB_DISABLE_XET=1 and removing any hf-xet or hf_transfer packages, the library continues attempting to contact cas-bridge.xethub.hf.co, which results in a repeated “RuntimeError: Data processing error: CAS service error : ReqwestMiddleware Error: Request failed after 5 retries”
HF_HUB_DISABLE_XET, HF_HUB_ENABLE_HF_TRANSFER)huggingface_hub to versions like 0.21.4, 0.23.0, and 0.30.2hf-xet is not installed (pip list, !find ~/.cache -name 'xet')hf_hub_download as well — same issuehf-xet to latest version - still the same errorThis turned out to be the ip whitelisting issue. After getting the below urls whitelisted, the model download worked with xet.
+" +404 Existing Hugging Face Inference Model Not Found,https://discuss.huggingface.co/t/404-existing-hugging-face-inference-model-not-found/165198,165198,23,2025-07-31 17:20:25.091000+00:00,"[{'id': 235857, 'name': 'Nolan Idle', 'username': 'AstroydsChat', 'avatar_template': '/user_avatar/discuss.huggingface.co/astroydschat/{size}/51945_2.png', 'created_at': '2025-07-31T17:20:25.147Z', 'cooked': 'So I am using the hugging face inference API and the model wont work on the inference API but works in the hugging face model playground: huggingface_hub.errors.HfHubHTTPError: 404 Client Error: Not Found for url: https://router.huggingface.co/hf-inference/models/HuggingFaceTB/SmolLM3-3B What should I do?
\nA more experience hugging face hub user.
\nMy own modified scripts
\nTo reproduce use the hugging face API on: HuggingFaceTB/SmolLM3-3B
\nThe expected behavior is to get a response to the request. When you get a parameter wrong when sending a request it gives a correct error message for that param but when you get everything correct it sends 404
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-31T17:20:25.147Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 109, 'reads': 13, 'readers_count': 12, 'score': 542.4, 'yours': False, 'topic_id': 165198, 'topic_slug': '404-existing-hugging-face-inference-model-not-found', 'display_username': 'Nolan Idle', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://router.huggingface.co/hf-inference/models/HuggingFaceTB/SmolLM3-3B', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100740, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/404-existing-hugging-face-inference-model-not-found/165198/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 235888, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-31T23:07:19.116Z', 'cooked': 'Hmm… Weird… It works with Python even without token…
\nimport os\nfrom huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n provider=""hf-inference"",\n #api_key=os.getenv(""HF_TOKEN"", None),\n)\n\ncompletion = client.chat.completions.create(\n model=""HuggingFaceTB/SmolLM3-3B"",\n messages=[\n {\n ""role"": ""user"",\n ""content"": ""What is the capital of France?""\n }\n ],\n)\n\nprint(completion.choices[0].message)\n#ChatCompletionOutputMessage(role=\'assistant\', content=""<think>\\nOkay, the user is asking for the capital of France. Let me make sure I remember correctly. I think it\'s Paris. Wait, is there any chance they might be confusing it with another city? Maybe they heard something different before?\\n\\nLet me double-check. France\'s capital is definitely Paris. It\'s the largest city in the country and a major cultural and political center. I don\'t think there\'s any other city that\'s considered the capital. Sometimes people might confuse it with Lyon or Marseille, but those are major cities too, not the capital.\\n\\nWait, what about the administrative capital? Oh right, even though Paris is the capital, some might refer to the administrative center as Paris as well. There\'s the Élysée Palace, which is the official residence of the President of France, and the seat of government. So yes, Paris is the capital.\\n\\nI should also consider if there\'s any historical context where another city might have been the capital. For example, during the French Revolution, Paris was the revolutionary capital, but it\'s still the capital now. There\'s no other city that\'s taken over as the capital in recent times.\\n\\nSo, the answer is Paris. I can confidently say that without any doubt. The user probably just needs a straightforward answer, but maybe they want a bit more context. I can mention that Paris is not only the capital but also a major cultural and economic hub in Europe. That adds value to the answer.\\n</think>\\n\\nThe capital of France is **Paris**. It is the largest city in the country and a prominent cultural, economic, and political center. Paris is known for iconic landmarks like the Eiffel Tower, the Louvre Museum, and Notre-Dame Cathedral."", tool_call_id=None, tool_calls=[], reasoning_content=None)\nHow about like this?
\ncurl -H ""Authorization: Bearer $HF_TOKEN"" \\\n https://router.huggingface.co/hf-inference/models/HuggingFaceTB/SmolLM3-3B\nSimilar issues:
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-31T23:23:56.213Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 10, 'readers_count': 9, 'score': 16.8, 'yours': False, 'topic_id': 165198, 'topic_slug': '404-existing-hugging-face-inference-model-not-found', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/issues/38524', 'internal': False, 'reflection': False, 'title': '404 Client Error when accessing https://router.huggingface.co/nebius/v1/chat/completions endpoint · Issue #38524 · huggingface/transformers · GitHub', 'clicks': 13}, {'url': 'https://github.com/huggingface/transformers/issues/39650', 'internal': False, 'reflection': False, 'title': 'Inference API Returning 404 · Issue #39650 · huggingface/transformers · GitHub', 'clicks': 11}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/404-existing-hugging-face-inference-model-not-found/165198/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 236162, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-08-02T16:19:43.596Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-08-02T16:19:43.596Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.2, 'yours': False, 'topic_id': 165198, 'topic_slug': '404-existing-hugging-face-inference-model-not-found', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/404-existing-hugging-face-inference-model-not-found/165198/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","So I am using the hugging face inference API and the model wont work on the inference API but works in the hugging face model playground: huggingface_hub.errors.HfHubHTTPError: 404 Client Error: Not Found for url: https://router.huggingface.co/hf-inference/models/HuggingFaceTB/SmolLM3-3B What should I do?
+A more experience hugging face hub user.
+My own modified scripts
+To reproduce use the hugging face API on: HuggingFaceTB/SmolLM3-3B
+The expected behavior is to get a response to the request. When you get a parameter wrong when sending a request it gives a correct error message for that param but when you get everything correct it sends 404
","Hmm… Weird… It works with Python even without token…
+import os
+from huggingface_hub import InferenceClient
+
+client = InferenceClient(
+ provider=""hf-inference"",
+ #api_key=os.getenv(""HF_TOKEN"", None),
+)
+
+completion = client.chat.completions.create(
+ model=""HuggingFaceTB/SmolLM3-3B"",
+ messages=[
+ {
+ ""role"": ""user"",
+ ""content"": ""What is the capital of France?""
+ }
+ ],
+)
+
+print(completion.choices[0].message)
+#ChatCompletionOutputMessage(role='assistant', content=""<think>\nOkay, the user is asking for the capital of France. Let me make sure I remember correctly. I think it's Paris. Wait, is there any chance they might be confusing it with another city? Maybe they heard something different before?\n\nLet me double-check. France's capital is definitely Paris. It's the largest city in the country and a major cultural and political center. I don't think there's any other city that's considered the capital. Sometimes people might confuse it with Lyon or Marseille, but those are major cities too, not the capital.\n\nWait, what about the administrative capital? Oh right, even though Paris is the capital, some might refer to the administrative center as Paris as well. There's the Élysée Palace, which is the official residence of the President of France, and the seat of government. So yes, Paris is the capital.\n\nI should also consider if there's any historical context where another city might have been the capital. For example, during the French Revolution, Paris was the revolutionary capital, but it's still the capital now. There's no other city that's taken over as the capital in recent times.\n\nSo, the answer is Paris. I can confidently say that without any doubt. The user probably just needs a straightforward answer, but maybe they want a bit more context. I can mention that Paris is not only the capital but also a major cultural and economic hub in Europe. That adds value to the answer.\n</think>\n\nThe capital of France is **Paris**. It is the largest city in the country and a prominent cultural, economic, and political center. Paris is known for iconic landmarks like the Eiffel Tower, the Louvre Museum, and Notre-Dame Cathedral."", tool_call_id=None, tool_calls=[], reasoning_content=None)
+How about like this?
+curl -H ""Authorization: Bearer $HF_TOKEN"" \
+ https://router.huggingface.co/hf-inference/models/HuggingFaceTB/SmolLM3-3B
+Similar issues:
+ +" +Spaces not working after restart,https://discuss.huggingface.co/t/spaces-not-working-after-restart/164981,164981,24,2025-07-29 17:09:44.710000+00:00,"[{'id': 235560, 'name': 'ezzdev', 'username': 'ezzdev', 'avatar_template': '/user_avatar/discuss.huggingface.co/ezzdev/{size}/31348_2.png', 'created_at': '2025-07-29T17:09:44.786Z', 'cooked': 'can you help me please solve this ?
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-07-29T17:09:44.786Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 6, 'readers_count': 5, 'score': 76.2, 'yours': False, 'topic_id': 164981, 'topic_slug': 'spaces-not-working-after-restart', 'display_username': 'ezzdev', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 63846, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-not-working-after-restart/164981/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 235565, 'name': 'ezzdev', 'username': 'ezzdev', 'avatar_template': '/user_avatar/discuss.huggingface.co/ezzdev/{size}/31348_2.png', 'created_at': '2025-07-29T17:31:06.032Z', 'cooked': 'i can confirm this happens only when using ZeroGPU but it works if i am using a paid GPU
\nAny help on this please ?
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-07-29T17:31:18.848Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 26.2, 'yours': False, 'topic_id': 164981, 'topic_slug': 'spaces-not-working-after-restart', 'display_username': 'ezzdev', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 63846, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-not-working-after-restart/164981/2', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235579, 'name': 'Saptarshi Neil Sinha', 'username': 'saptarshineilsinha', 'avatar_template': '/user_avatar/discuss.huggingface.co/saptarshineilsinha/{size}/51857_2.png', 'created_at': '2025-07-29T18:32:29.409Z', 'cooked': 'Same issue from myside
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-07-29T18:32:29.409Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 164981, 'topic_slug': 'spaces-not-working-after-restart', 'display_username': 'Saptarshi Neil Sinha', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 63846, 'username': 'ezzdev', 'name': 'ezzdev', 'avatar_template': '/user_avatar/discuss.huggingface.co/ezzdev/{size}/31348_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100578, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-not-working-after-restart/164981/3', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235581, 'name': 'Saptarshi Neil Sinha', 'username': 'saptarshineilsinha', 'avatar_template': '/user_avatar/discuss.huggingface.co/saptarshineilsinha/{size}/51857_2.png', 'created_at': '2025-07-29T18:33:40.197Z', 'cooked': 'Seems to be working with only CPU but not zeroGPU : On restart ZeroGPU not working but on CPU it works
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-07-29T18:33:40.197Z', 'reply_count': 1, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 31.0, 'yours': False, 'topic_id': 164981, 'topic_slug': 'spaces-not-working-after-restart', 'display_username': 'Saptarshi Neil Sinha', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/on-restart-zerogpu-not-working-but-on-cpu-it-works/164979', 'internal': True, 'reflection': False, 'title': 'On restart ZeroGPU not working but on CPU it works', 'clicks': 2}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 100578, 'username': 'saptarshineilsinha', 'name': 'Saptarshi Neil Sinha', 'avatar_template': '/user_avatar/discuss.huggingface.co/saptarshineilsinha/{size}/51857_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100578, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-not-working-after-restart/164981/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235584, 'name': 'ezzdev', 'username': 'ezzdev', 'avatar_template': '/user_avatar/discuss.huggingface.co/ezzdev/{size}/31348_2.png', 'created_at': '2025-07-29T18:52:07.402Z', 'cooked': 'the issue solved after restart and factory rebuild
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-07-29T18:52:07.402Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 164981, 'topic_slug': 'spaces-not-working-after-restart', 'display_username': 'ezzdev', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 100578, 'username': 'saptarshineilsinha', 'name': 'Saptarshi Neil Sinha', 'avatar_template': '/user_avatar/discuss.huggingface.co/saptarshineilsinha/{size}/51857_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 63846, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-not-working-after-restart/164981/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 235641, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-30T06:52:21.658Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-07-30T06:52:21.658Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 164981, 'topic_slug': 'spaces-not-working-after-restart', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/spaces-not-working-after-restart/164981/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","can you help me please solve this ?
",the issue solved after restart and factory rebuild
+Inference providers: Access to processor data?,https://discuss.huggingface.co/t/inference-providers-access-to-processor-data/164824,164824,64,2025-07-28 15:49:02.752000+00:00,"[{'id': 235357, 'name': 'Frank Sommers', 'username': 'fsommers', 'avatar_template': '/user_avatar/discuss.huggingface.co/fsommers/{size}/36212_2.png', 'created_at': '2025-07-28T15:49:02.812Z', 'cooked': 'I love the HF inference providers, but now ran into a question:
\nIs it possible to get access to the model’s processor output as well via the API?
\nMy specific use-case is with Qwen2.5-VL. I ask the model to perform localization tasks on document images. I ask the model to find bounding box coordinates for page elements. The model generally does very well in this task.
\nIn order to correctly map the localization data returned from the model to my original image sizes, I found that I needed to access the processor’s inputs. That’s because the Qwen processor adjusts image sizes, something that I think is pretty common for many models working with vision encoders. In my case, using the transformers library:
\ninputs = processor(text=[text], images=images, padding=True, return_tensors=""pt"")\n...\noutput_ids = model.generate(**inputs, max_new_tokens=max_new_tokens)\ngenerated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(inputs.input_ids, output_ids)]\noutput_text = processor.batch_decode(generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True)\n \n# Now I can obtain the input image size:\ninput_height = inputs[\'image_grid_thw\'][0][1]*14\ninput_width = inputs[\'image_grid_thw\'][0][2]*14\nThe model’s localization coordinates will be based on that image size, and this is important to scale those coordinates to some other image dimensions the user actually sees.
\nHow could I solve this using the Inference API?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-28T15:50:35.364Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 6, 'readers_count': 5, 'score': 36.2, 'yours': False, 'topic_id': 164824, 'topic_slug': 'inference-providers-access-to-processor-data', 'display_username': 'Frank Sommers', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 74253, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inference-providers-access-to-processor-data/164824/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 235422, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-29T00:50:43.329Z', 'cooked': 'If it were a Dedicated Endpoint that you could maintain yourself, you could change the return value by just rewriting handler.py, but since you are using the Inference Provider, that part is a black box.
Therefore, as you suggested, mimicking the processing that is likely being done internally is a relatively lightweight and better approach…
\nWith the following code, the entire model will not be downloaded. It should be possible to use JSON alone.
from PIL import Image\nimport requests\nfrom transformers import AutoProcessor\n\nurl = ""https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/diffusion-quicktour.png""\norig = Image.open(requests.get(url, stream=True).raw)\nprompt = ""describe this image""\nprocessor = AutoProcessor.from_pretrained(""Qwen/Qwen2.5-VL-7B-Instruct"")\n\ninputs = processor(images=[orig], text=[prompt], padding=True, return_tensors=""pt"")\n\ngrid_h, grid_w = inputs[""image_grid_thw""][0][1:].tolist()\nproc_h, proc_w = grid_h * 14, grid_w * 14\nsx, sy = orig.width / proc_w, orig.height / proc_h\nprint(inputs[""image_grid_thw""], sx, sy) # tensor([[ 1, 18, 18]]) 1.0158730158730158 1.0158730158730158\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-29T12:50:49.075Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 164824, 'topic_slug': 'inference-providers-access-to-processor-data', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/inference-providers-access-to-processor-data/164824/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I love the HF inference providers, but now ran into a question:
+Is it possible to get access to the model’s processor output as well via the API?
+My specific use-case is with Qwen2.5-VL. I ask the model to perform localization tasks on document images. I ask the model to find bounding box coordinates for page elements. The model generally does very well in this task.
+In order to correctly map the localization data returned from the model to my original image sizes, I found that I needed to access the processor’s inputs. That’s because the Qwen processor adjusts image sizes, something that I think is pretty common for many models working with vision encoders. In my case, using the transformers library:
+inputs = processor(text=[text], images=images, padding=True, return_tensors=""pt"")
+...
+output_ids = model.generate(**inputs, max_new_tokens=max_new_tokens)
+generated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(inputs.input_ids, output_ids)]
+output_text = processor.batch_decode(generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True)
+
+# Now I can obtain the input image size:
+input_height = inputs['image_grid_thw'][0][1]*14
+input_width = inputs['image_grid_thw'][0][2]*14
+The model’s localization coordinates will be based on that image size, and this is important to scale those coordinates to some other image dimensions the user actually sees.
+How could I solve this using the Inference API?
","If it were a Dedicated Endpoint that you could maintain yourself, you could change the return value by just rewriting handler.py, but since you are using the Inference Provider, that part is a black box.
Therefore, as you suggested, mimicking the processing that is likely being done internally is a relatively lightweight and better approach…
+With the following code, the entire model will not be downloaded. It should be possible to use JSON alone.
from PIL import Image
+import requests
+from transformers import AutoProcessor
+
+url = ""https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/diffusion-quicktour.png""
+orig = Image.open(requests.get(url, stream=True).raw)
+prompt = ""describe this image""
+processor = AutoProcessor.from_pretrained(""Qwen/Qwen2.5-VL-7B-Instruct"")
+
+inputs = processor(images=[orig], text=[prompt], padding=True, return_tensors=""pt"")
+
+grid_h, grid_w = inputs[""image_grid_thw""][0][1:].tolist()
+proc_h, proc_w = grid_h * 14, grid_w * 14
+sx, sy = orig.width / proc_w, orig.height / proc_h
+print(inputs[""image_grid_thw""], sx, sy) # tensor([[ 1, 18, 18]]) 1.0158730158730158 1.0158730158730158
+I am using AutoTrain to finetune my Llama model with my custom data and the model give random responses ignoring my dataset. The thing is that on my dataset I have 145 rows in JSONL and when I start the fine-tuning with this dataset and I analyze logs I can see these rows:
\n
\nSo the dataset is recognized with 145 rows so from here I can understand that my dataset is well-structured and every row is a valid JSON object.
\nBut right after the model shards are uploaded, it gives me this log:
Generating train split: 0 examples [00:00, ? examples/s]\nGenerating train split: 9 examples [00:00, ? examples/s]\nSo my question is: Why does it log Generating train split 0 examples and Generating train split 9 examples right below?
\nIs this a normal behaviour of AutoTrain?
\nOr there’s something that I have to adjust on my training dataset?
\nAfter the model is finetuned, obviously I can see it on my HuggingFace hub and I can also see the training statistics on TensorBoard but I see only a dot on the graphs and the training loss about 5.4, so yeah, everytime I try to ask him something about my dataset or anything else, he answers me randomly.
\nWhat can I do in order to finetune a model in the right way? Maybe I just have to expand my dataset because 145 rows are not enough and those logs are just normal?
\n\nWhy does it log Generating train split 0 examples and Generating train split 9 examples right below?
\n
This error seems to occur when Column Mapping is not set correctly.
', 'post_number': 2, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-28T10:01:44.837Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 7, 'readers_count': 6, 'score': 11.4, 'yours': False, 'topic_id': 164782, 'topic_slug': 'model-responses-are-random-ignoring-my-dataset', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/autotrain/col_map', 'internal': False, 'reflection': False, 'title': 'Understanding Column Mapping', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-responses-are-random-ignoring-my-dataset/164782/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235314, 'name': 'Matthias Di Giorgio', 'username': 'DigioMatthy', 'avatar_template': '/user_avatar/discuss.huggingface.co/digiomatthy/{size}/51793_2.png', 'created_at': '2025-07-28T10:18:12.332Z', 'cooked': 'My dataset is a jsonl format and has only one column ‘text’.
\nIn AutoTrain I set the Column Mapping like this:
\n
It appears to be correct… Another possible factor is that packing is enabled with the small dataset.
\nAlso, unless there is a specific reason, I think it’s safer to leave Chat Template on automatic.
Following the general documentation on the Column Mapping in AutoTrain topic I tried to set the Column Mapping like this:
\n
So now looking at the discussion they are talking about disabling the parameter packing but the thing is that even if I enable full parameter mode there is no packing parameter, anyway I’m using basic parameter mode because otherwise I don’t know what to tweak.
\nMaybe do I have to write manually parameters activating JSON parameters first and doing so I can write like packing=false and try with other parameters?
\nOr maybe it’s just my dataset too small and I have to expand it?
There is no doubt that the dataset is too small, but I don’t think it’s absolutely impossible with that amount of data…
\nIf there is a publicly available dataset that can reproduce the symptoms, it would be possible to investigate…
\nIf there are no settings for packing, it will be difficult with SFT with small dataset…
', 'post_number': 6, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-28T13:03:22.896Z', 'reply_count': 2, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 11.0, 'yours': False, 'topic_id': 164782, 'topic_slug': 'model-responses-are-random-ignoring-my-dataset', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/trl/en/sft_trainer#packing-dataset', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-responses-are-random-ignoring-my-dataset/164782/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235333, 'name': 'Matthias Di Giorgio', 'username': 'DigioMatthy', 'avatar_template': '/user_avatar/discuss.huggingface.co/digiomatthy/{size}/51793_2.png', 'created_at': '2025-07-28T13:22:04.588Z', 'cooked': 'Ok it was predictable that the dataset was too small for a real fine-tuning actually, I’ll create a bigger one and I’ll try launch a finetuning and we’ll see if I will have the same problem, but I don’t think so .
\nLast question, what do you think the minimal amount of examples a dataset should have in order to make a really good and successful fine-tuning?
Ah I forgot to say, maybe the issue could be that AutoTrain GUI doesn’t permit to set a value to a packing parameter because behind it’s a default set and it can’t be handled, so if someone wants to train their own model, the dataset has to be large
', 'post_number': 8, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-28T13:26:55.111Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 164782, 'topic_slug': 'model-responses-are-random-ignoring-my-dataset', 'display_username': 'Matthias Di Giorgio', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100457, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-responses-are-random-ignoring-my-dataset/164782/8', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235339, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-28T13:49:26.532Z', 'cooked': 'Hmm, I think you should ask someone who knows more about LLM fine-tuning than I do, but what I sometimes hear is that “500 to 1000 samples are sufficient for LoRA”, “data diversity is more important than quantity”, etc.
\nSince it is difficult to manually create a dataset from scratch, many people choose to use existing AI tools to create dataset. Also, the online documents like this may be useful references regarding formatting.
', 'post_number': 9, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-28T13:49:26.532Z', 'reply_count': 1, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 164782, 'topic_slug': 'model-responses-are-random-ignoring-my-dataset', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://arxiv.org/pdf/2305.11206', 'internal': False, 'reflection': False, 'clicks': 1}, {'url': 'https://huggingface.co/blog/tegridydev/llm-dataset-formats-101-hugging-face', 'internal': False, 'reflection': False, 'title': 'LLM Dataset Formats 101: A No‐BS Guide for Hugging Face Devs', 'clicks': 0}, {'url': 'https://huggingface.co/posts/CultriX/959128360368232', 'internal': False, 'reflection': False, 'title': '@CultriX on Hugging Face: ""Script for QA-style dataset generation from custom data: Transform Your…""', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 100457, 'username': 'DigioMatthy', 'name': 'Matthias Di Giorgio', 'avatar_template': '/user_avatar/discuss.huggingface.co/digiomatthy/{size}/51793_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-responses-are-random-ignoring-my-dataset/164782/9', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 235341, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-28T13:55:12.236Z', 'cooked': 'There are people who know more about AI than I do who say things like, “Ask AI about AI.” Commercial AI systems like Gemini and ChatGPT have been trained on a lot of AI-related information, so when you ask them about AI itself, they often provide fairly reliable answers. Since they have a solid foundation of knowledge, even just enabling search can help you gather reasonably up-to-date information.
', 'post_number': 10, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-28T13:55:12.236Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 164782, 'topic_slug': 'model-responses-are-random-ignoring-my-dataset', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-responses-are-random-ignoring-my-dataset/164782/10', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235342, 'name': 'Matthias Di Giorgio', 'username': 'DigioMatthy', 'avatar_template': '/user_avatar/discuss.huggingface.co/digiomatthy/{size}/51793_2.png', 'created_at': '2025-07-28T13:55:18.161Z', 'cooked': 'Ok, I think these documentations you pinged me are enough to solve the dataset problem.
\nThank you so much for your time and support!!
Wow, didn’t know that. Ok will try it then! Ty!!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 13, 'post_type': 3, 'posts_count': 13, 'updated_at': '2025-07-29T01:56:48.470Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 164782, 'topic_slug': 'model-responses-are-random-ignoring-my-dataset', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/model-responses-are-random-ignoring-my-dataset/164782/13', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am using AutoTrain to finetune my Llama model with my custom data and the model give random responses ignoring my dataset. The thing is that on my dataset I have 145 rows in JSONL and when I start the fine-tuning with this dataset and I analyze logs I can see these rows:
+
+So the dataset is recognized with 145 rows so from here I can understand that my dataset is well-structured and every row is a valid JSON object.
+But right after the model shards are uploaded, it gives me this log:
Generating train split: 0 examples [00:00, ? examples/s]
+Generating train split: 9 examples [00:00, ? examples/s]
+So my question is: Why does it log Generating train split 0 examples and Generating train split 9 examples right below?
+Is this a normal behaviour of AutoTrain?
+Or there’s something that I have to adjust on my training dataset?
+After the model is finetuned, obviously I can see it on my HuggingFace hub and I can also see the training statistics on TensorBoard but I see only a dot on the graphs and the training loss about 5.4, so yeah, everytime I try to ask him something about my dataset or anything else, he answers me randomly.
+What can I do in order to finetune a model in the right way? Maybe I just have to expand my dataset because 145 rows are not enough and those logs are just normal?
Hmm, I think you should ask someone who knows more about LLM fine-tuning than I do, but what I sometimes hear is that “500 to 1000 samples are sufficient for LoRA”, “data diversity is more important than quantity”, etc.
+Since it is difficult to manually create a dataset from scratch, many people choose to use existing AI tools to create dataset. Also, the online documents like this may be useful references regarding formatting.
" +How to save my model to use it later,https://discuss.huggingface.co/t/how-to-save-my-model-to-use-it-later/20568,20568,5,2022-07-19 12:37:44.659000+00:00,"[{'id': 40527, 'name': 'Hoss', 'username': 'slowturtle', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/77aa72/{size}.png', 'created_at': '2022-07-19T12:37:44.729Z', 'cooked': 'Hello Amazing people,
\nThis is my first post and I am really new to machine learning and Hugginface.
I followed this awesome guide here multilabel Classification with DistilBert
\nand used my dataset and the results are very good. I am having a hard time know trying to understand how to save the model I trainned and all the artifacts needed to use my model later.
\nI tried at the end of the tutorial: torch.save(trainer, \'my_model\') but I got this error msg:
AttributeError: Can\'t pickle local object \'get_linear_schedule_with_warmup.<locals>.lr_lambda\'
I have the following files saved for each epoch:
\nconfig.json\n optimizer.pt\n pytorch_model.bin\n rng_state.pth\n special_tokens_map.json\n tokenizer.json\n tokenizer_config.json\n trainer_state.json\n training_args.bin\n vocab.txt\nCan someone kindly guide me how to save this model to later use?
\nThank you very much
Hello there,
\nYou can save models with trainer.save_model(""path_to_save""). Another cool thing you can do is you can push your model to the Hugging Face Hub as well. I added couple of lines to notebook to show you, here. You can find pushing there.
Thank you very much for helping me Merve. Huge Thanks.
\nJust one more question if you don’t mind: I’ll now use my model locally at first. You helped me to save all the files I need to load it again.
So to use the same model I save with trainer.save_model(path) I just need to use trainer.load(path)?
Thank you very much
Hello again,
\nYou can simply load the model using the model class’ from_pretrained(model_path) method like below:
\n(you can either save locally and load from local or push to Hub and load from Hub)
from transformers import BertConfig, BertModel\n# if model is on hugging face Hub\nmodel = BertModel.from_pretrained(""bert-base-uncased"")\n# from local folder\nmodel = BertModel.from_pretrained(""./test/saved_model/"")\nAnother cool thing you can use is pipeline API, it will make your life much easier . With pipelines, you will not have to deal with internals of the model or tokenizer to infer with the model, you simply give the folder and it will make the model ready to infer for you.
You are amazing merve I’ll try do to this steps now. Let’s see how it goes.
\nThank you again
Hello again,
\nSo I followed that tutorial to train my model(using distilert-base-uncased).
\nsaved the model with:
trainer.save_model(""./my_model"")
and then I loaded the model:
\nfrom transformers import DistilBertConfig, DistilBertModel\npath = \'path_to_my_model\'\nmodel = DistilBertModel.from_pretrained(path)\nNow I followed the same tutorial for inference but then I run:
\nencoding = tokenizer(text, return_tensors=""pt"")\n\nencoding = {k: v.to(trainer.model.device) for k,v in encoding.items()}\noutputs = trainer.model(**encoding)\nand then:
\nlogits = outputs.logits raises the followin error:
AttributeError: \'DistilBertModel\' object has no attribute \'logits\'
How can I fix this step?
\nThank you very much
', 'post_number': 6, 'post_type': 1, 'posts_count': 18, 'updated_at': '2022-07-19T16:31:23.749Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 2006, 'reads': 2286, 'readers_count': 2285, 'score': 10507.2, 'yours': False, 'topic_id': 20568, 'topic_slug': 'how-to-save-my-model-to-use-it-later', 'display_username': 'Hoss', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 4339, 'username': 'merve', 'name': 'merve', 'avatar_template': '/user_avatar/discuss.huggingface.co/merve/{size}/49809_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8979, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-save-my-model-to-use-it-later/20568/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 40589, 'name': 'Hoss', 'username': 'slowturtle', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/77aa72/{size}.png', 'created_at': '2022-07-19T21:52:50.489Z', 'cooked': 'I found the error: instead of
\nmodel = DistilBertModel.from_pretrained(path)
\nI changed to
\nmodel = AutoModelForSequenceClassification.from_pretrained(path)
@slowturtle Just to avoid confusion for future, the BertModel classes are simply BERT models without classification heads on top, so the heads include classification heads (and thus logit processors).
', 'post_number': 8, 'post_type': 1, 'posts_count': 18, 'updated_at': '2022-07-20T11:24:09.025Z', 'reply_count': 0, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 283, 'reads': 1697, 'readers_count': 1696, 'score': 1769.4, 'yours': False, 'topic_id': 20568, 'topic_slug': 'how-to-save-my-model-to-use-it-later', 'display_username': 'merve', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 8979, 'username': 'slowturtle', 'name': 'Hoss', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/77aa72/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': True, 'staff': True, 'user_id': 4339, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-save-my-model-to-use-it-later/20568/8', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 53183, 'name': 'Ishan Babbar', 'username': 'ishan42d', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/i/c6cbf5/{size}.png', 'created_at': '2022-12-28T00:21:34.670Z', 'cooked': 'Hi Merve!
\nI might be late but the tutorial that you have shared is excellent. My only questions is that can the same model be trained for a Multiclass text classification problem as well? If so, what parameters do I need to keep in mind while training this model? and also will this be successful for smaller datasets (<1000 records). It will be great to see if you have a notebook for this problem statement as well that I have just described
\nThanks
\nIshan
Hi!
\nI run out of CUDA memory when saving a larger model using this. Is there a way I can move a gpu trained model to ‘cpu’ before saving using trainer.save_model(_). Appreciate the help, thanks!
', 'post_number': 10, 'post_type': 1, 'posts_count': 18, 'updated_at': '2023-06-15T15:24:52.362Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 228, 'reads': 1044, 'readers_count': 1043, 'score': 1368.8, 'yours': False, 'topic_id': 20568, 'topic_slug': 'how-to-save-my-model-to-use-it-later', 'display_username': 'Naman ', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 4339, 'username': 'merve', 'name': 'merve', 'avatar_template': '/user_avatar/discuss.huggingface.co/merve/{size}/49809_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 22130, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-save-my-model-to-use-it-later/20568/10', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 86945, 'name': 'Nikos Peppes', 'username': 'nikospps', 'avatar_template': '/user_avatar/discuss.huggingface.co/nikospps/{size}/19016_2.png', 'created_at': '2023-08-30T13:33:17.991Z', 'cooked': 'Hello. After running a distilbert model, finetuned with my own custom dataset for classification purposes, i try to save the model in a .pth file format (e.g. distilmodel.pth). After training the model using the Trainer from the pytorch library, it saves a couples of archives into a checkpoint output folder, as declared into the Trainer’s arguments.
\nAny help to convert the checkpoint into a model.pth format file?
\nThanks in advance.
What if we want to take a base model from HuggingFace, train it, save the fine-tune model, and then train it further? I want to train the model iteratively on subsets of my data so I don’t have to train it all at once because it will take a few weeks to do it all at once and I am afraid it will crash towards the end and waste the experiment, as well as I want to be able to test the output in between subsets of data.
\nCurrently, when I try to load a custom model and tokenizer, though I can generate text with the model no problem, I get the below error when I attempt to train it further:
\nExpected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument mat2 in method wrapper_CUDA_mm)\nThe thing is, this is not an issue when I train the base model model initially, but I have even tried forcing the data to be on the GPU before training and then just get the same error complaining about cuda:0 and cuda:3. I think the data moves to the GPU after training.Train() is called, and all my settings are the same besides the fact I am referencing my locally saved model and tokenizer path instead of the HuggingFace web path. Do I need to push my model to huggingface and then download from there? I looked at the folders that are cached from downloading the model and there are quite a few extra files that are cached aside from the files created when I save the model to a local folder, but any help would be very appreciated.
', 'post_number': 12, 'post_type': 1, 'posts_count': 18, 'updated_at': '2023-12-26T19:29:18.858Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 116, 'reads': 599, 'readers_count': 598, 'score': 699.8, 'yours': False, 'topic_id': 20568, 'topic_slug': 'how-to-save-my-model-to-use-it-later', 'display_username': 'Ryan Farran', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31398, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-save-my-model-to-use-it-later/20568/12', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 108363, 'name': 'Akindele Michael', 'username': 'DeleMike', 'avatar_template': '/user_avatar/discuss.huggingface.co/delemike/{size}/26732_2.png', 'created_at': '2024-01-14T21:38:48.982Z', 'cooked': '\n\nI am using this repo to run a translation task. Especially I’m using it to build a diacritization model. I need to save the model after the process is done. How can I do that?
\nCUDA_VISIBLE_DEVICES=0 python run_translation.py --model_name_or_path Davlan/oyo-t5-small --do_train --do_eval --source_lang unyo --target_lang dcyo --source_prefix ""<unyo2dcyo>: "" --train_file data_prep_eng/output_data/bible_train.json --validation_file data_prep_eng/output_data/dev.json --test_file data_prep_eng/output_data/test.json --output_dir oyot5_small_unyo_dcyo_bible --max_source_length 512 --max_target_length 512 --per_device_train_batch_size=24 --per_device_eval_batch_size=24 --num_train_epochs 3 --overwrite_output_dir --predict_with_generate --save_steps 10000 --num_beams 10 --do_predict \nAm I missing a flag like --save-model? I need the saved model to be part of the directory.
See what I have now:
\n
Yes, you can. Assuming you are using torch:
\nDEVICE = “cpu”
\n#assuming huggingface model
\nyour_model.to(DEVICE)
GPU_DEVICE = “cuda” if torch.cuda.is_available() else “cpu”
', 'post_number': 14, 'post_type': 1, 'posts_count': 18, 'updated_at': '2024-01-26T05:57:26.991Z', 'reply_count': 0, 'reply_to_post_number': 10, 'quote_count': 0, 'incoming_link_count': 105, 'reads': 367, 'readers_count': 366, 'score': 598.4, 'yours': False, 'topic_id': 20568, 'topic_slug': 'how-to-save-my-model-to-use-it-later', 'display_username': 'Cybrtooth', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 22130, 'username': 'naman-trilogy', 'name': 'Naman ', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/n/eb8c5e/{size}.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 37195, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-save-my-model-to-use-it-later/20568/14', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 115453, 'name': 'Yaoming Xuan', 'username': 'Greykxu', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/g/bbce88/{size}.png', 'created_at': '2024-02-23T10:49:17.739Z', 'cooked': 'Hi, thanks for the answer. But is there a method or convention to NOT use trainer to save models?
\nI prefer to finetune my model by training in the traditional pytorch way because it’s more flexiable to add my own creativity. But I find it difficult to save it. The error message says that I shouldn’t use the identical checkpointing as the original model. What does it mean? Is there any method to solve it?
how to save dreams on huggingface and on the blockchain ? You may think i am a dreamer but i am not the only one - Research - Hugging Face Forums
', 'post_number': 16, 'post_type': 1, 'posts_count': 18, 'updated_at': '2024-11-10T04:30:56.724Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 37, 'reads': 81, 'readers_count': 80, 'score': 201.2, 'yours': False, 'topic_id': 20568, 'topic_slug': 'how-to-save-my-model-to-use-it-later', 'display_username': None, 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 70114, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-save-my-model-to-use-it-later/20568/16', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 235014, 'name': 'Mohamed Gomaa', 'username': 'Coalbbb', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/ccd318/{size}.png', 'created_at': '2025-07-26T09:29:10.469Z', 'cooked': 'I have a question about saving models. If I use model.save_pretrained(), will it save the original weights that weren’t optimized during training?
models/, checkpoints/)cnn_cifar10_2025-07-28_acc93.h5Hello Amazing people,
+This is my first post and I am really new to machine learning and Hugginface.
I followed this awesome guide here multilabel Classification with DistilBert
+and used my dataset and the results are very good. I am having a hard time know trying to understand how to save the model I trainned and all the artifacts needed to use my model later.
+I tried at the end of the tutorial: torch.save(trainer, 'my_model') but I got this error msg:
AttributeError: Can't pickle local object 'get_linear_schedule_with_warmup.<locals>.lr_lambda'
I have the following files saved for each epoch:
+config.json
+ optimizer.pt
+ pytorch_model.bin
+ rng_state.pth
+ special_tokens_map.json
+ tokenizer.json
+ tokenizer_config.json
+ trainer_state.json
+ training_args.bin
+ vocab.txt
+Can someone kindly guide me how to save this model to later use?
+Thank you very much
Hello again,
+So I followed that tutorial to train my model(using distilert-base-uncased).
+saved the model with:
trainer.save_model(""./my_model"")
and then I loaded the model:
+from transformers import DistilBertConfig, DistilBertModel
+path = 'path_to_my_model'
+model = DistilBertModel.from_pretrained(path)
+Now I followed the same tutorial for inference but then I run:
+encoding = tokenizer(text, return_tensors=""pt"")
+
+encoding = {k: v.to(trainer.model.device) for k,v in encoding.items()}
+outputs = trainer.model(**encoding)
+and then:
+logits = outputs.logits raises the followin error:
AttributeError: 'DistilBertModel' object has no attribute 'logits'
How can I fix this step?
+Thank you very much
" +Fine-tune Mistral 7B–9B or 24B (bnb 4bit),https://discuss.huggingface.co/t/fine-tune-mistral-7b-9b-or-24b-bnb-4bit/164597,164597,9,2025-07-26 12:47:57.932000+00:00,"[{'id': 235043, 'name': 'Nikita', 'username': 'oukaise', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/a87d85/{size}.png', 'created_at': '2025-07-26T12:47:57.990Z', 'cooked': 'Hi everyone,
\nI’m exploring the feasibility of fine-tuning a 7B–9B model (like Mistral or Deepseek) on consumer hardware using 4-bit quantization (bnb). My current setup:
\nSpecs:
\nUse case:
\nI’m building a system that generates short, contextualized outputs based on external content. The goal is to make the model more domain-aware by giving it access to a corpus of ~9k domain-specific text entries (no outputs), and then fine-tune it to better generate responses when paired with smaller adapters (LoRAs) per user or use-case (each around 200–300 examples).
Pipeline idea:
\nMy questions:
\nCan Mistral 7B or Deepseek 9B (bnb-4bit) be fine-tuned efficiently on V100 16GB using tools like Unsloth?
\nIf I add a second GPU (e.g. another V100, P100 16GB, or RTX 3060 12GB), is it possible to:
\nWhat’s the recommended approach for managing 10+ LoRAs for runtime personalization?
\nWhich models are generally best suited for this kind of task (short domain-aware output generation + user-specific fine-tuning)?
\nI’m currently looking at Mistral, Deepseek, Yi, LLaMA 3, but open to suggestions for 4-bit setups on limited VRAM.
Any practical insights, configs, or success stories would be super appreciated!
\nThanks a lot.
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-26T13:45:29.205Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 83, 'reads': 4, 'readers_count': 3, 'score': 415.8, 'yours': False, 'topic_id': 164597, 'topic_slug': 'fine-tune-mistral-7b-9b-or-24b-bnb-4bit', 'display_username': 'Nikita', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100356, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fine-tune-mistral-7b-9b-or-24b-bnb-4bit/164597/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 235046, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-26T13:47:56.461Z', 'cooked': 'For now, with 24B seems difficult with just one card, but with 7B should be doable.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-26T13:47:56.461Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 164597, 'topic_slug': 'fine-tune-mistral-7b-9b-or-24b-bnb-4bit', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://www.unsloth.ai/blog/mistral-small-3.1', 'internal': False, 'reflection': False, 'title': 'Fine-tune Mistral Small 3.1 with Unsloth', 'clicks': 4}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fine-tune-mistral-7b-9b-or-24b-bnb-4bit/164597/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 235054, 'name': 'Nikita', 'username': 'oukaise', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/a87d85/{size}.png', 'created_at': '2025-07-26T15:07:04.780Z', 'cooked': 'what if i use two gpus
\nlike two v100s with 16gb
\nor a v100 + p100 16gb
\nor rtx 3060 12gb + v100
\nbut most likely just for inference, and for full fine-tuning i’d rent a server for 2–3 days and then use the result
\nwould that work?
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-27T03:07:57.243Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 164597, 'topic_slug': 'fine-tune-mistral-7b-9b-or-24b-bnb-4bit', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/fine-tune-mistral-7b-9b-or-24b-bnb-4bit/164597/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I’m exploring the feasibility of fine-tuning a 7B–9B model (like Mistral or Deepseek) on consumer hardware using 4-bit quantization (bnb). My current setup:
+Specs:
+Use case:
+I’m building a system that generates short, contextualized outputs based on external content. The goal is to make the model more domain-aware by giving it access to a corpus of ~9k domain-specific text entries (no outputs), and then fine-tune it to better generate responses when paired with smaller adapters (LoRAs) per user or use-case (each around 200–300 examples).
Pipeline idea:
+My questions:
+Can Mistral 7B or Deepseek 9B (bnb-4bit) be fine-tuned efficiently on V100 16GB using tools like Unsloth?
+If I add a second GPU (e.g. another V100, P100 16GB, or RTX 3060 12GB), is it possible to:
+What’s the recommended approach for managing 10+ LoRAs for runtime personalization?
+Which models are generally best suited for this kind of task (short domain-aware output generation + user-specific fine-tuning)?
+I’m currently looking at Mistral, Deepseek, Yi, LLaMA 3, but open to suggestions for 4-bit setups on limited VRAM.
Any practical insights, configs, or success stories would be super appreciated!
+Thanks a lot.
","For now, with 24B seems difficult with just one card, but with 7B should be doable.
" +Trainer never invokes compute_metrics,https://discuss.huggingface.co/t/trainer-never-invokes-compute-metrics/11440,11440,5,2021-11-07 21:55:35.715000+00:00,"[{'id': 24642, 'name': 'bnqu', 'username': 'nbqu', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/n/9dc877/{size}.png', 'created_at': '2021-11-07T21:55:35.796Z', 'cooked': 'def compute_metrics(p: EvalPrediction):\n print(""***Computing Metrics***"") # THIS LINE NEVER PRINTED\n preds = p.predictions[0] if isinstance(p.predictions, tuple) else p.predictions\n preds = np.squeeze(preds) if is_regression else np.argmax(preds, axis=1)\n if data_args.task_name is not None:\n result = metric.compute(predictions=preds, references=p.label_ids)\n if len(result) > 1:\n result[""combined_score""] = np.mean(list(result.values())).item()\n return result\n elif is_regression:\n return {""mse"": ((preds - p.label_ids) ** 2).mean().item()}\n else:\n return {""accuracy"": (preds == p.label_ids).astype(np.float32).mean().item()}\n\n...\n\n # Initialize our Trainer\n trainer = Trainer(\n model=model,\n args=training_args,\n train_dataset=train_dataset if training_args.do_train else None,\n eval_dataset=eval_dataset if training_args.do_eval else None,\n compute_metrics=compute_metrics,\n tokenizer=tokenizer,\n data_collator=data_collator,\n )\n\n # Training\n if training_args.do_train:\n checkpoint = None\n if training_args.resume_from_checkpoint is not None:\n checkpoint = training_args.resume_from_checkpoint\n elif last_checkpoint is not None:\n checkpoint = last_checkpoint\n train_result = trainer.train(resume_from_checkpoint=checkpoint)\n metrics = train_result.metrics\n max_train_samples = (\n data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset)\n )\n metrics[""train_samples""] = min(max_train_samples, len(train_dataset))\n\n trainer.save_model() # Saves the tokenizer too for easy upload\n trainer.log_metrics(""train"", metrics)\n trainer.save_metrics(""train"", metrics)\n trainer.save_state()\n\n if training_args.do_eval:\n logger.info(""*** Evaluate ***"")\n\n # Loop to handle MNLI double evaluation (matched, mis-matched)\n tasks = [data_args.task_name]\n eval_datasets = [eval_dataset]\n if data_args.task_name == ""mnli"":\n tasks.append(""mnli-mm"")\n eval_datasets.append(raw_datasets[""validation_mismatched""])\n\n for eval_dataset, task in zip(eval_datasets, tasks):\n metrics = trainer.evaluate(eval_dataset=eval_dataset)\n\n max_eval_samples = (\n data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset)\n )\n metrics[""eval_samples""] = min(max_eval_samples, len(eval_dataset))\n\n trainer.log_metrics(""eval"", metrics)\n trainer.save_metrics(""eval"", metrics)\n ""output_dir"": ""./output_dir"",\n ""do_train"": true,\n ""do_eval"": true,\n ""learning_rate"": 1e-5,\n ""per_device_train_batch_size"": 32,\n ""per_device_eval_batch_size"": 32,\n ""logging_strategy"": ""epoch"",\n ""save_strategy"": ""epoch"",\n ""evaluation_strategy"": ""epoch"",\n ""prediction_loss_only"": false,\nI have a question during training my own dataset, forked base code from run_glue.py. The arguments are my TrainingArguments.
\nDuring training / validation, it seems that compute_metrics never invoked while other things run correctly.
How can I fix this so I can get accuracy or other metrics?
\nPlease let me know if you need more information or code
Are you sure your datasets has proper labels? This may be the reason the compute metrics is skipped.
', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2021-11-08T13:08:14.302Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 46, 'reads': 264, 'readers_count': 263, 'score': 287.8, 'yours': False, 'topic_id': 11440, 'topic_slug': 'trainer-never-invokes-compute-metrics', 'display_username': 'Sylvain Gugger', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/trainer-never-invokes-compute-metrics/11440/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 24720, 'name': 'bnqu', 'username': 'nbqu', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/n/9dc877/{size}.png', 'created_at': '2021-11-09T00:34:38.685Z', 'cooked': 'Hi, I investigated the code with debugger,
\nand I checked whether there is labels before I put my eval_dataset (in case of evaluation) to trainer.evaluate(). code example
I got batched eval_dataset with shape (batch_size, 6) which is consist of
\n[\'attention_mask\', \'input_ids\', \'label\', \'sentence1\', \'sentence2\', \'token_type_ids\'], and there were proper labels as you concerned.
Is there any ways to get access inside of the inner method evaluation_loop so I can check how it works?
You can see the batches that will be passed to your model for evaluation with:
\nfor batch in trainer.get_eval_dataloader(eval_dataset):\n break\nAnd see if it does contain the ""labels"" key.
\'labels\' key in batch but still Trainer doesn’t return metrics.\nI would just return to classic and compute metrics manually for now…
\nThank you for your answer!
Hi,
\nI have the same problem and it still does not work
for batch in trainer.get_eval_dataloader(eval_dataset):\n print(batch)\n break\ngives me “labels” but the compute_metrics function is never called. What else has to be configures ?
\nthanks !
Avoid modifying TrainingArguments keys manually, especially for the evaluation strategy, logging strategy or save strategy. Indeed the __post_init__ from TrainingArguments makes sure we use instances of IntervalStrategy and not simple strings, so if you override with e.g. training_args.evaluation_strategy = ""steps"" you will have troubles. If you really need to override, use training_args.evaluation_strategy = IntervalStrategy.STEPS
See transformers/trainer_callback.py at 8afaaa26f5754948f4ddf8f31d70d0293488a897 · huggingface/transformers · GitHub and transformers/training_args.py at 8afaaa26f5754948f4ddf8f31d70d0293488a897 · huggingface/transformers · GitHub
', 'post_number': 7, 'post_type': 1, 'posts_count': 8, 'updated_at': '2022-04-26T14:51:52.428Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 128, 'reads': 186, 'readers_count': 185, 'score': 707.2, 'yours': False, 'topic_id': 11440, 'topic_slug': 'trainer-never-invokes-compute-metrics', 'display_username': 'Félix Marty', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/blob/8afaaa26f5754948f4ddf8f31d70d0293488a897/src/transformers/trainer_callback.py#L420', 'internal': False, 'reflection': False, 'title': 'transformers/trainer_callback.py at 8afaaa26f5754948f4ddf8f31d70d0293488a897 · huggingface/transformers · GitHub', 'clicks': 174}, {'url': 'https://github.com/huggingface/transformers/blob/8afaaa26f5754948f4ddf8f31d70d0293488a897/src/transformers/training_args.py#L804', 'internal': False, 'reflection': False, 'title': 'transformers/training_args.py at 8afaaa26f5754948f4ddf8f31d70d0293488a897 · huggingface/transformers · GitHub', 'clicks': 108}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 7404, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/trainer-never-invokes-compute-metrics/11440/7', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234890, 'name': 'Hugo Fara', 'username': 'hugofara', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/h/e36b37/{size}.png', 'created_at': '2025-07-25T08:45:35.964Z', 'cooked': 'I had the same issue.
\nMy problem was that I was compute_loss_func in TrainingArgs, instead of defining it from inside the model. It prevents the evaluation function to run.
def compute_metrics(p: EvalPrediction):
+ print(""***Computing Metrics***"") # THIS LINE NEVER PRINTED
+ preds = p.predictions[0] if isinstance(p.predictions, tuple) else p.predictions
+ preds = np.squeeze(preds) if is_regression else np.argmax(preds, axis=1)
+ if data_args.task_name is not None:
+ result = metric.compute(predictions=preds, references=p.label_ids)
+ if len(result) > 1:
+ result[""combined_score""] = np.mean(list(result.values())).item()
+ return result
+ elif is_regression:
+ return {""mse"": ((preds - p.label_ids) ** 2).mean().item()}
+ else:
+ return {""accuracy"": (preds == p.label_ids).astype(np.float32).mean().item()}
+
+...
+
+ # Initialize our Trainer
+ trainer = Trainer(
+ model=model,
+ args=training_args,
+ train_dataset=train_dataset if training_args.do_train else None,
+ eval_dataset=eval_dataset if training_args.do_eval else None,
+ compute_metrics=compute_metrics,
+ tokenizer=tokenizer,
+ data_collator=data_collator,
+ )
+
+ # Training
+ if training_args.do_train:
+ checkpoint = None
+ if training_args.resume_from_checkpoint is not None:
+ checkpoint = training_args.resume_from_checkpoint
+ elif last_checkpoint is not None:
+ checkpoint = last_checkpoint
+ train_result = trainer.train(resume_from_checkpoint=checkpoint)
+ metrics = train_result.metrics
+ max_train_samples = (
+ data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset)
+ )
+ metrics[""train_samples""] = min(max_train_samples, len(train_dataset))
+
+ trainer.save_model() # Saves the tokenizer too for easy upload
+ trainer.log_metrics(""train"", metrics)
+ trainer.save_metrics(""train"", metrics)
+ trainer.save_state()
+
+ if training_args.do_eval:
+ logger.info(""*** Evaluate ***"")
+
+ # Loop to handle MNLI double evaluation (matched, mis-matched)
+ tasks = [data_args.task_name]
+ eval_datasets = [eval_dataset]
+ if data_args.task_name == ""mnli"":
+ tasks.append(""mnli-mm"")
+ eval_datasets.append(raw_datasets[""validation_mismatched""])
+
+ for eval_dataset, task in zip(eval_datasets, tasks):
+ metrics = trainer.evaluate(eval_dataset=eval_dataset)
+
+ max_eval_samples = (
+ data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset)
+ )
+ metrics[""eval_samples""] = min(max_eval_samples, len(eval_dataset))
+
+ trainer.log_metrics(""eval"", metrics)
+ trainer.save_metrics(""eval"", metrics)
+ ""output_dir"": ""./output_dir"",
+ ""do_train"": true,
+ ""do_eval"": true,
+ ""learning_rate"": 1e-5,
+ ""per_device_train_batch_size"": 32,
+ ""per_device_eval_batch_size"": 32,
+ ""logging_strategy"": ""epoch"",
+ ""save_strategy"": ""epoch"",
+ ""evaluation_strategy"": ""epoch"",
+ ""prediction_loss_only"": false,
+I have a question during training my own dataset, forked base code from run_glue.py. The arguments are my TrainingArguments.
+During training / validation, it seems that compute_metrics never invoked while other things run correctly.
How can I fix this so I can get accuracy or other metrics?
+Please let me know if you need more information or code
You can see the batches that will be passed to your model for evaluation with:
+for batch in trainer.get_eval_dataloader(eval_dataset):
+ break
+And see if it does contain the ""labels"" key.
Hey guys
\nI’m struggling with this error:
\n404 Client Error: Not Found for url: https://router.huggingface.co/hf-inference/models/Qwen/Qwen2.5-Coder-32B-Instruct/v1/chat/completions
\nThe code is taken from here:
\n\n\nIt’s appearing with any instruct model i tried (including those with special access such as Llama models)
\nWhat’s that?
\nWould be grateful for any help
\nI saw there is maybe a problem with zero-scale or something like that, but i used popular models, I’m not sure that this is a reason
', 'post_number': 1, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-12T11:58:39.553Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 683, 'reads': 32, 'readers_count': 31, 'score': 2965.6, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'Alevtina Vesper', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/learn/agents-course/en/unit2/llama-index/llama-hub', 'internal': False, 'reflection': False, 'clicks': 10}, {'url': 'https://router.huggingface.co/hf-inference/models/Qwen/Qwen2.5-Coder-32B-Instruct/v1/chat/completions', 'internal': False, 'reflection': False, 'clicks': 3}, {'url': 'https://discuss.huggingface.co/t/api-access-disabled/164844/2', 'internal': True, 'reflection': True, 'title': 'API Access Disabled?', 'clicks': 3}, {'url': 'https://discuss.huggingface.co/t/agents-course-unit-2-2-error-404/168035/4', 'internal': True, 'reflection': True, 'title': 'Agents Course Unit 2.2 error 404', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/serverless-inference-api-always-returns-404-even-for-public-models/166845/2', 'internal': True, 'reflection': True, 'title': 'Serverless Inference API always returns 404, even for public models', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/error-401-client-error-unauthorized-for-url/19714/79', 'internal': True, 'reflection': True, 'title': 'Error 401 Client Error: Unauthorized for url', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/api-returns-not-found-invalid-credentials-for-any-key-from-new-verified-accounts/163823/2', 'internal': True, 'reflection': True, 'title': 'API returns ""Not Found"" / ""Invalid Credentials"" for any key from new verified accounts', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 232413, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-12T12:40:21.292Z', 'cooked': 'I think this is due to a large number of models whose deployment has been canceled, as well as major changes to the library used for the Inference API. I’m not familiar with the workaround for this issue on LlamaIndex, but according to GitHub, updating the HF library should still make it work.
\npip install -U huggingface_hub\nHi, thanks for your answer!
\nUnfortunately updating didn’t help, I’ve tried it
Hmm, in that case, do you need to update LlamaIndex, or has it become unusable due to further specification changes…?
\nI think the model itself is deployed via Inference Provider.
However, if you are not particularly attached to that model, it might be better to look for an alternative. More detailed information is available in the Agents course channel on Hugging Face Discord.
\nEverything is up-to-date
\nActually I’m using some other models directly, but just want to cope with that problem. Maybe someone knows how to fix it
\nThank you anyway
', 'post_number': 5, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-12T14:57:28.982Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 19, 'readers_count': 18, 'score': 33.4, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'Alevtina Vesper', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 232471, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-12T23:17:25.884Z', 'cooked': '\n\n\n
\n\nhf-inference <= this
\n
I see. Let me explain the situation. It is normal for this URL not to work because this model has not been deployed with HF Inference. Currently, very few LLMs are deployed via HF Inference. Most are deployed via other Inference Providers.
\nIf LlamaIndex does not have a feature to switch the Inference Provider or set it to =""auto"", only few models will work.
Yes, I think you’re right and the problem is in the framework or so. Just don’t understand why they put this example in the course.
\nActually it must be available for deploy with HF Inference, because there is a code for deploying:
import os\nfrom huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n provider=""auto"",\n api_key=os.environ[""HF_TOKEN""],\n)\n\ncompletion = client.chat.completions.create(\n model=""Qwen/Qwen2.5-Coder-32B-Instruct"",\n messages=[\n {\n ""role"": ""user"",\n ""content"": ""What is the capital of France?""\n }\n ],\n)\n\nprint(completion.choices[0].message)\nBut maybe this is the only way to deploy it, and HuggingFaceInferenceAPI is restricted now (despite this code is in the course).
', 'post_number': 7, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-13T05:01:13.343Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 18, 'readers_count': 17, 'score': 28.2, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'Alevtina Vesper', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 232504, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-13T05:06:13.541Z', 'cooked': '\n\nJust don’t understand why they put this example in the course.
\n
Yeah. When the course was created, that method was available…
\nIf it’s just a matter of library versions or so, we can just stick with the old ones, but for the “Agents” course, we need as many examples as possible of using “external APIs,” whether provided by HF or a third party…
But AI services change a lot in just a few months. It’s difficult to keep them up to date.
\n', 'post_number': 8, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-13T05:06:13.541Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 21, 'readers_count': 20, 'score': 28.8, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/agents-course/issues', 'internal': False, 'reflection': False, 'title': 'GitHub · Where software is built', 'clicks': 7}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/8', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 232507, 'name': 'Alevtina Vesper', 'username': 'TinaVesper', 'avatar_template': '/user_avatar/discuss.huggingface.co/tinavesper/{size}/50991_2.png', 'created_at': '2025-07-13T05:36:01.590Z', 'cooked': 'Agree. But it can be easily resolved at least with linked discussions about problems&solutions on this forum for instance. Just one button on the page “Got stuck, but found a solution? Tell us more” or so. I saw the same on the other platform. Or just a little checklist, like..there are may appear some problems. Check you have Pro status to use HF Inference API, check deploy button etc etc
\nWithout claims to authors, always there are ways to make a course better
\nThanks for you help!
', 'post_number': 9, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-13T05:38:13.029Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 21, 'readers_count': 20, 'score': 58.8, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'Alevtina Vesper', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234668, 'name': 'Dzung Le', 'username': 'dzungever', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/46a35a/{size}.png', 'created_at': '2025-07-24T05:36:18.602Z', 'cooked': 'I can get HuggingFaceInferenceAPI to work by adding the provider as below.
\nllm = HuggingFaceInferenceAPI(
\nmodel_name=“Qwen/Qwen2.5-Coder-32B-Instruct”,
\ntemperature=0.7,
\nmax_tokens=100,
\ntoken=hf_token,
\nprovider=“together”,
\n)
Hmm, that’s strange… I think it’s been deployed…
\nHave you tried updating LangChain and huggingface_hub?
Edit:
\nOh. I misunderstood. Great!
\nMaybe provider=""auto"", also work.
Yes, it works this way, thanks a lot!
', 'post_number': 12, 'post_type': 1, 'posts_count': 13, 'updated_at': '2025-07-24T06:18:01.918Z', 'reply_count': 0, 'reply_to_post_number': 10, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 19, 'readers_count': 18, 'score': 48.4, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'Alevtina Vesper', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 96595, 'username': 'dzungever', 'name': 'Dzung Le', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/46a35a/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99241, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/12', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234803, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-24T18:18:59.504Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 13, 'post_type': 3, 'posts_count': 13, 'updated_at': '2025-07-24T18:18:59.504Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 11, 'readers_count': 10, 'score': 1.8, 'yours': False, 'topic_id': 162747, 'topic_slug': 'hf-agents-course-404-client-error-not-found-for-url', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/hf-agents-course-404-client-error-not-found-for-url/162747/13', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hey guys
+I’m struggling with this error:
+404 Client Error: Not Found for url: https://router.huggingface.co/hf-inference/models/Qwen/Qwen2.5-Coder-32B-Instruct/v1/chat/completions
+The code is taken from here:
+ + +It’s appearing with any instruct model i tried (including those with special access such as Llama models)
+What’s that?
+Would be grateful for any help
+I saw there is maybe a problem with zero-scale or something like that, but i used popular models, I’m not sure that this is a reason
","I can get HuggingFaceInferenceAPI to work by adding the provider as below.
+llm = HuggingFaceInferenceAPI(
+model_name=“Qwen/Qwen2.5-Coder-32B-Instruct”,
+temperature=0.7,
+max_tokens=100,
+token=hf_token,
+provider=“together”,
+)
Hello,
\nI am getting a persistent 401 Unauthorized error in Google Colab when trying to download any gated model, such as meta-llama/Meta-Llama-3-8B-Instruct.
I have already confirmed on the model’s webpage that I have been granted access. The error continues even after I generate a brand new write token and pass it directly to the from_pretrained function in my code.
This suggests a possible issue with my account’s token validation, as all standard debugging steps have failed. Could you please advise?
\nThank you.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-19T23:19:50.363Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 10, 'reads': 4, 'readers_count': 3, 'score': 60.8, 'yours': False, 'topic_id': 163756, 'topic_slug': 'persistent-401-unauthorized-error-on-gated-models', 'display_username': 'Alvin Siphosenkosi Moyo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99812, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/persistent-401-unauthorized-error-on-gated-models/163756/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 233917, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-20T02:22:29.743Z', 'cooked': 'First, try whoami-v2, which should make verification easy.
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-20T02:22:29.743Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 163756, 'topic_slug': 'persistent-401-unauthorized-error-on-gated-models', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/how-do-you-use-the-whoami-endpoint/15830/2', 'internal': True, 'reflection': False, 'title': 'How do you use the whoami endpoint?', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/persistent-401-unauthorized-error-on-gated-models/163756/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233982, 'name': 'Alvin Siphosenkosi Moyo', 'username': 'AlvinSiphosenkosi', 'avatar_template': '/user_avatar/discuss.huggingface.co/alvinsiphosenkosi/{size}/51382_2.png', 'created_at': '2025-07-20T13:57:07.918Z', 'cooked': 'Hello,
\nFollowing up on my 401 Unauthorized issue. I have run the command-line diagnostic tool as requested.
When I run huggingface-cli whoami, I get the following explicit error:
Invalid user token. The token from HF_TOKEN environment variable is invalid.{""error"":""Invalid credentials in Authorization header""}
I have meticulously regenerated and pasted a new write token multiple times, and the error persists. This definitively proves the problem is not with my code but with the token validation for my account. Can you please investigate the status of my account and tokens?
Thank you.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-20T13:57:07.918Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 163756, 'topic_slug': 'persistent-401-unauthorized-error-on-gated-models', 'display_username': 'Alvin Siphosenkosi Moyo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99812, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/persistent-401-unauthorized-error-on-gated-models/163756/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233984, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-20T14:09:24.257Z', 'cooked': 'If the problem is account-specific, I think it would be quicker to contact Hugging Face support. website@huggingface.co
\nAnother case that occasionally occurs is that extra information is added when copying and pasting tokens. This is more likely to happen when using shortcut keys.
\nIn addition, there are many conditions that cause a 401 error.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-07-24T11:12:19.399Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 163756, 'topic_slug': 'persistent-401-unauthorized-error-on-gated-models', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/persistent-401-unauthorized-error-on-gated-models/163756/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+I am getting a persistent 401 Unauthorized error in Google Colab when trying to download any gated model, such as meta-llama/Meta-Llama-3-8B-Instruct.
I have already confirmed on the model’s webpage that I have been granted access. The error continues even after I generate a brand new write token and pass it directly to the from_pretrained function in my code.
This suggests a possible issue with my account’s token validation, as all standard debugging steps have failed. Could you please advise?
+Thank you.
","If the problem is account-specific, I think it would be quicker to contact Hugging Face support. website@huggingface.co
+Another case that occasionally occurs is that extra information is added when copying and pasting tokens. This is more likely to happen when using shortcut keys.
+In addition, there are many conditions that cause a 401 error.
This link works Exp - a Hugging Face Space by user93729
\nBut this link gives 404 https://user93729-exp.hf.space/
\nIt is a static HTML page. Why doesn’t the direct link work?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-23T01:30:35.726Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 44, 'reads': 8, 'readers_count': 7, 'score': 236.6, 'yours': False, 'topic_id': 164180, 'topic_slug': 'static-html-space-direct-link-gives-404', 'display_username': 'User 93729', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/user93729/exp', 'internal': False, 'reflection': False, 'title': 'Exp - a Hugging Face Space by user93729', 'clicks': 2}, {'url': 'https://user93729-exp.hf.space/', 'internal': False, 'reflection': False, 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 100078, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/static-html-space-direct-link-gives-404/164180/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 234462, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-23T02:20:25.446Z', 'cooked': 'In static space, seems the URL will be like this.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-23T02:20:25.446Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 16.6, 'yours': False, 'topic_id': 164180, 'topic_slug': 'static-html-space-direct-link-gives-404', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://user93729-exp.static.hf.space', 'internal': False, 'reflection': False, 'title': 'KCl Detector Count Calculator', 'clicks': 3}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/static-html-space-direct-link-gives-404/164180/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 234465, 'name': 'izum00', 'username': 'soiz1', 'avatar_template': '/user_avatar/discuss.huggingface.co/soiz1/{size}/51492_2.png', 'created_at': '2025-07-23T02:35:12.803Z', 'cooked': ' user93729-exp.hf.space/index.html
\n user93729-exp. static. hf.space/index.html
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-23T14:35:44.637Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 164180, 'topic_slug': 'static-html-space-direct-link-gives-404', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/static-html-space-direct-link-gives-404/164180/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","This link works Exp - a Hugging Face Space by user93729
+But this link gives 404 https://user93729-exp.hf.space/
+It is a static HTML page. Why doesn’t the direct link work?
","In static space, seems the URL will be like this.
" +Dataset scripts are no longer supported,https://discuss.huggingface.co/t/dataset-scripts-are-no-longer-supported/163891,163891,10,2025-07-21 04:59:31.021000+00:00,"[{'id': 234067, 'name': 'kajal gupta', 'username': 'kajalhappy', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/k/54ee81/{size}.png', 'created_at': '2025-07-21T04:59:31.085Z', 'cooked': 'I was using earlier GeneratorBasedBuilder class for loading database now i am getting below error :
\nException occurred: Dataset scripts are no longer supported.
using load_dataset for loading script . but somehow it is not supported not
\nplz tell me what is the other way to load database using GeneratorBasedBuilder class i need to pre process the database before saving in arrow or other format.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-21T05:28:33.025Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3937, 'reads': 55, 'readers_count': 54, 'score': 18100.2, 'yours': False, 'topic_id': 163891, 'topic_slug': 'dataset-scripts-are-no-longer-supported', 'display_username': 'kajal gupta', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 35652, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-scripts-are-no-longer-supported/163891/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 234081, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-21T07:36:00.171Z', 'cooked': 'Seems trust_remote_code is deprecated in datasets 4.0.0.
\nSo quick workarounds:
pip install datasets<4.0.0 \nIn addition, it seems that downgrading huggingface_hub may be necessary in some cases.
I clarify this just in case. It seems that support for the function to build datasets locally would continue.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-21T12:10:26.545Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 33, 'reads': 52, 'readers_count': 51, 'score': 159.8, 'yours': False, 'topic_id': 163891, 'topic_slug': 'dataset-scripts-are-no-longer-supported', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/datasets/pull/7592#issuecomment-3079918731', 'internal': False, 'reflection': False, 'title': 'Remove scripts altogether by lhoestq · Pull Request #7592 · huggingface/datasets · GitHub', 'clicks': 155}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-scripts-are-no-longer-supported/163891/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234240, 'name': 'kajal gupta', 'username': 'kajalhappy', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/k/54ee81/{size}.png', 'created_at': '2025-07-22T04:53:17.587Z', 'cooked': 'yes, we can not use load_dataset if implementing a Builder class.
\nso need to explicitly call builder class and generate dataset.
\nbuilder.download_and_prepare()
\ndataset = builder.as_dataset(split=Split.TRAIN)
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-07-22T16:53:47.183Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 49, 'reads': 38, 'readers_count': 37, 'score': 242.2, 'yours': False, 'topic_id': 163891, 'topic_slug': 'dataset-scripts-are-no-longer-supported', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/dataset-scripts-are-no-longer-supported/163891/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I was using earlier GeneratorBasedBuilder class for loading database now i am getting below error :
+Exception occurred: Dataset scripts are no longer supported.
using load_dataset for loading script . but somehow it is not supported not
+plz tell me what is the other way to load database using GeneratorBasedBuilder class i need to pre process the database before saving in arrow or other format.
","yes, we can not use load_dataset if implementing a Builder class.
+so need to explicitly call builder class and generate dataset.
+builder.download_and_prepare()
+dataset = builder.as_dataset(split=Split.TRAIN)
I am trying to use the facebook/data2vec-audio-base-960h model.
\nAs per their model card, this is how to load the model:
from transformers import Wav2Vec2Processor, Data2VecForCTC\n\n processor = Wav2Vec2Processor.from_pretrained(""facebook/data2vec-audio-base-960h"")\n model = Data2VecForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")\nBut I am getting this error:
\nImportError Traceback (most recent call last)\n/tmp/ipython-input-11-2185350118.py in <cell line: 0>()\n----> 1 from transformers import Wav2Vec2Processor, Data2VecForCTC\n 2 \n 3 processor = Wav2Vec2Processor.from_pretrained(""facebook/data2vec-audio-base-960h"")\n 4 model = Data2VecForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")\n\nImportError: cannot import name \'Wav2Vec2Processor\' from \'transformers\' (/usr/local/lib/python3.11/dist-packages/transformers/__init__.py)\nI looked up at stack-overflow: It suggested upgrading the Transformers version.
\nI did that :
It seems that the previous sample on the web was incorrect, and now it works on my Colab.
\n!pip install -U transformers accelerate huggingface_hub[hf_xet]\n\n#from transformers import Wav2Vec2Processor, Data2VecForCTC\nfrom transformers import Wav2Vec2Processor, Data2VecAudioForCTC\n\nprocessor = Wav2Vec2Processor.from_pretrained(""facebook/data2vec-audio-base-960h"")\n#model = Data2VecForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")\nmodel = Data2VecAudioForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-22T14:08:56.176Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 163992, 'topic_slug': 'cannot-import-name-wav2vec2processor', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cannot-import-name-wav2vec2processor/163992/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to use the facebook/data2vec-audio-base-960h model.
+As per their model card, this is how to load the model:
from transformers import Wav2Vec2Processor, Data2VecForCTC
+
+ processor = Wav2Vec2Processor.from_pretrained(""facebook/data2vec-audio-base-960h"")
+ model = Data2VecForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")
+But I am getting this error:
+ImportError Traceback (most recent call last)
+/tmp/ipython-input-11-2185350118.py in <cell line: 0>()
+----> 1 from transformers import Wav2Vec2Processor, Data2VecForCTC
+ 2
+ 3 processor = Wav2Vec2Processor.from_pretrained(""facebook/data2vec-audio-base-960h"")
+ 4 model = Data2VecForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")
+
+ImportError: cannot import name 'Wav2Vec2Processor' from 'transformers' (/usr/local/lib/python3.11/dist-packages/transformers/__init__.py)
+I looked up at stack-overflow: It suggested upgrading the Transformers version.
+I did that :
It seems that the previous sample on the web was incorrect, and now it works on my Colab.
+!pip install -U transformers accelerate huggingface_hub[hf_xet]
+
+#from transformers import Wav2Vec2Processor, Data2VecForCTC
+from transformers import Wav2Vec2Processor, Data2VecAudioForCTC
+
+processor = Wav2Vec2Processor.from_pretrained(""facebook/data2vec-audio-base-960h"")
+#model = Data2VecForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")
+model = Data2VecAudioForCTC.from_pretrained(""facebook/data2vec-audio-base-960h"")
+I run below code on a RTX 3090 with Ryzen 9 7900X and 128 GB RAM. So generating a single 512x512 image takes 20 minutes.
\nIs that normal? I read that it just should take seconds.
import torch\nfrom diffusers import FluxPipeline\nimport sys\nimport time\n\nstart = time.time()\nprint(""CUDA available:"", torch.cuda.is_available())\nprint(""Device:"", torch.cuda.get_device_name(0) if torch.cuda.is_available() else ""CPU"")\n\npipe = FluxPipeline.from_pretrained(""black-forest-labs/FLUX.1-dev"", torch_dtype=torch.bfloat16)\npipe.to(""cuda"")\n\nprompt = ""a wolf running""\n\nimages_ = pipe(\n prompt,\n # width=1920,\n # height=1088,\n width=512,\n height=512,\n guidance_scale=3.5,\n num_inference_steps=50,\n max_sequence_length=512,\n generator=torch.Generator(device=""cuda"").manual_seed(0)\n).images\n\nfor i, image in enumerate(images_):\n image.save(""flux-dev"" + str(i) + "".png"")\n\nend = time.time()\nprint(f""Generation took {time.time() - start:.2f} seconds"")\nCuda is 12.1, PYthon is 3.10
\nPackages (installed version | lastest version):
| GitPython | \n3.1.44 | \n3.1.44 | \n
|---|---|---|
| MarkupSafe | \n2.1.5 | \n3.0.2 | \n
| PyYAML | \n6.0.2 | \n6.0.2 | \n
| accelerate | \n1.9.0 | \n1.9.0 | \n
| aiofiles | \n23.2.1 | \n24.1.0 | \n
| altair | \n5.5.0 | \n5.5.0 | \n
| annotated-types | \n0.7.0 | \n0.7.0 | \n
| anyio | \n4.9.0 | \n4.9.0 | \n
| attrs | \n25.3.0 | \n25.3.0 | \n
| blinker | \n1.9.0 | \n1.9.0 | \n
| cachetools | \n6.1.0 | \n6.1.0 | \n
| certifi | \n2025.7.14 | \n2025.7.14 | \n
| charset-normalizer | \n3.4.2 | \n3.4.2 | \n
| click | \n8.2.1 | \n8.2.1 | \n
| colorama | \n0.4.6 | \n0.4.6 | \n
| diffusers | \n0.34.0 | \n0.34.0 | \n
| einops | \n0.8.1 | \n0.8.1 | \n
| exceptiongroup | \n1.3.0 | \n1.3.0 | \n
| fastapi | \n0.116.1 | \n0.116.1 | \n
| ffmpy | \n0.6.0 | \n0.6.0 | \n
| filelock | \n3.18.0 | \n3.18.0 | \n
| fire | \n0.7.0 | \n0.7.0 | \n
| flux | \n0.0.post58+g1371b2b | \n1.3.5 | \n
| fsspec | \n2025.7.0 | \n2025.7.0 | \n
| gitdb | \n4.0.12 | \n4.0.12 | \n
| gradio | \n5.13.2 | \n5.38.0 | \n
| gradio-client | \n1.6.0 | \n1.11.0 | \n
| h11 | \n0.16.0 | \n0.16.0 | \n
| httpcore | \n1.0.9 | \n1.0.9 | \n
| httpx | \n0.28.1 | \n0.28.1 | \n
| huggingface-hub | \n0.33.4 | \n0.33.4 | \n
| idna | \n3.10 | \n3.10 | \n
| importlib-metadata | \n8.7.0 | \n8.7.0 | \n
| invisible-watermark | \n0.2.0 | \n0.2.0 | \n
| jinja2 | \n3.1.6 | \n3.1.6 | \n
| jsonschema | \n4.25.0 | \n4.25.0 | \n
| jsonschema-specifications | \n2025.4.1 | \n2025.4.1 | \n
| markdown-it-py | \n3.0.0 | \n3.0.0 | \n
| mdurl | \n0.1.2 | \n0.1.2 | \n
| mpmath | \n1.3.0 | \n1.3.0 | \n
| narwhals | \n1.48.0 | \n1.48.0 | \n
| networkx | \n3.4.2 | \n3.5 | \n
| numpy | \n2.2.6 | \n2.3.1 | \n
| opencv-python | \n4.12.0.88 | \n4.12.0.88 | \n
| orjson | \n3.11.0 | \n3.11.0 | \n
| packaging | \n25.0 | \n25.0 | \n
| pandas | \n2.3.1 | \n2.3.1 | \n
| pillow | \n11.3.0 | \n11.3.0 | \n
| pip | \n25.1.1 | \n25.1.1 | \n
| protobuf | \n6.31.1 | \n6.31.1 | \n
| psutil | \n7.0.0 | \n7.0.0 | \n
| pyarrow | \n21.0.0 | \n21.0.0 | \n
| pydantic | \n2.11.7 | \n2.11.7 | \n
| pydantic-core | \n2.33.2 | \n\n |
| pydeck | \n0.9.1 | \n0.9.1 | \n
| pydub | \n0.25.1 | \n0.25.1 | \n
| pygments | \n2.19.2 | \n2.19.2 | \n
| python-dateutil | \n2.9.0.post0 | \n2.9.0.post0 | \n
| python-multipart | \n0.0.20 | \n0.0.20 | \n
| pytz | \n2025.2 | \n2025.2 | \n
| pywavelets | \n1.8.0 | \n1.8.0 | \n
| referencing | \n0.36.2 | \n0.36.2 | \n
| regex | \n2024.11.6 | \n2024.11.6 | \n
| requests | \n2.32.4 | \n2.32.4 | \n
| rich | \n14.0.0 | \n14.0.0 | \n
| rpds-py | \n0.26.0 | \n0.26.0 | \n
| ruff | \n0.6.8 | \n0.12.4 | \n
| safehttpx | \n0.1.6 | \n0.1.6 | \n
| safetensors | \n0.5.3 | \n0.5.3 | \n
| semantic-version | \n2.10.0 | \n2.10.0 | \n
| sentencepiece | \n0.2.0 | \n0.2.0 | \n
| setuptools | \n57.4.0 | \n80.9.0 | \n
| shellingham | \n1.5.4 | \n1.5.4 | \n
| six | \n1.17.0 | \n1.17.0 | \n
| smmap | \n5.0.2 | \n6.0.0 | \n
| sniffio | \n1.3.1 | \n1.3.1 | \n
| starlette | \n0.47.2 | \n0.47.2 | \n
| streamlit | \n1.47.0 | \n1.47.0 | \n
| streamlit-drawable-canvas | \n0.9.3 | \n0.9.3 | \n
| streamlit-keyup | \n0.3.0 | \n0.3.0 | \n
| sympy | \n1.13.1 | \n1.14.0 | \n
| tenacity | \n9.1.2 | \n9.1.2 | \n
| termcolor | \n3.1.0 | \n3.1.0 | \n
| tokenizers | \n0.21.2 | \n0.21.2 | \n
| toml | \n0.10.2 | \n0.10.2 | \n
| tomlkit | \n0.13.3 | \n0.13.3 | \n
| torch | \n2.5.1+cu121 | \n2.7.1 | \n
| torchaudio | \n2.5.1+cu121 | \n2.7.1 | \n
| torchvision | \n0.20.1+cu121 | \n0.22.1 | \n
| tornado | \n6.5.1 | \n6.5.1 | \n
| tqdm | \n4.67.1 | \n4.67.1 | \n
| transformers | \n4.53.2 | \n4.53.2 | \n
| typer | \n0.16.0 | \n0.16.0 | \n
| typing-extensions | \n4.14.1 | \n4.14.1 | \n
| typing-inspection | \n0.4.1 | \n0.4.1 | \n
| tzdata | \n2025.2 | \n2025.2 | \n
| urllib3 | \n2.5.0 | \n2.5.0 | \n
| uvicorn | \n0.35.0 | \n0.35.0 | \n
| watchdog | \n6.0.0 | \n6.0.0 | \n
| websockets | \n14.2 | \n15.0.1 | \n
| zipp | \n3.23.0 | \n3.23.0 | \n
\n\non a RTX 3090 with Ryzen 9 7900X and 128 GB RAM. So generating a single 512x512 image takes 20 minutes.
\n
\nIs that normal?
Yeah. With that code, FLUX is loaded into VRAM or RAM in a 16-bit state without quantization, requiring approximately 36 GB or more. Since VRAM is insufficient, it cannot be utilized effectively, resulting in lengthy inference times. Therefore,
\nQuantization is at least necessary. For 4-bit quantization methods, I recommend BitsAndBytes for ease of use or TorchAO for speed.
\nWhile there were various limitations when using LoRA in the past, these should be largely resolved now.
Optimization methods for FLUX:
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-21T11:50:18.479Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 163940, 'topic_slug': 'how-long-does-image-generation-with-black-forest-labs-flux-1-dev-take', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/diffusers/main/en/optimization/memory', 'internal': False, 'reflection': False, 'title': 'Reduce memory usage', 'clicks': 3}, {'url': 'https://huggingface.co/blog/diffusers-quantization', 'internal': False, 'reflection': False, 'title': 'Exploring Quantization Backends in Diffusers', 'clicks': 2}, {'url': 'https://pytorch.org/blog/torch-compile-and-diffusers-a-hands-on-guide-to-peak-performance/', 'internal': False, 'reflection': False, 'clicks': 0}, {'url': 'https://huggingface.co/docs/diffusers/main/en/optimization/para_attn', 'internal': False, 'reflection': False, 'title': 'ParaAttention', 'clicks': 0}, {'url': 'https://github.com/huggingface/diffusers/pull/9453', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-long-does-image-generation-with-black-forest-labs-flux-1-dev-take/163940/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 234174, 'name': 'Dent Black', 'username': 'RTQAQ', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/r/59ef9b/{size}.png', 'created_at': '2025-07-21T17:08:50.224Z', 'cooked': 'Thanks for the answer. I could reduce the runtime from 20 min to 2min.
\nDo you see any possible improvements with my code?
\nI adjusted the code to:
import torch\nfrom diffusers import FluxPipeline, DiffusionPipeline\nimport time, os\nfrom diffusers.quantizers import PipelineQuantizationConfig\nfrom datetime import datetime\n\nstart = time.time()\n\ntorch._dynamo.config.capture_dynamic_output_shape_ops = True\n\n# quantize\npipeline_quant_config = PipelineQuantizationConfig(\n quant_backend=""bitsandbytes_4bit"",\n quant_kwargs={""load_in_4bit"": True, ""bnb_4bit_quant_type"": ""nf4"", ""bnb_4bit_compute_dtype"": torch.bfloat16},\n components_to_quantize=[""transformer"", ""text_encoder_2""],\n)\npipeline = DiffusionPipeline.from_pretrained(\n ""black-forest-labs/FLUX.1-dev"",\n quantization_config=pipeline_quant_config,\n torch_dtype=torch.bfloat16,\n).to(""cuda"")\n\n# compile\npipeline.transformer.to(memory_format=torch.channels_last)\n\nprompt = ""a wolf running"" \n\nimages_ = pipeline(\n prompt,\n width=1920,\n height=1088,\n # width=64,\n # height=64,\n guidance_scale=3.5,\n num_inference_steps=50,\n max_sequence_length=512,\n generator=torch.Generator(device=""cuda"").manual_seed(0)).images\nThere are no major issues, so I think you can proceed by adding optimization methods based on that.
\nThe specific optimization methods available will vary depending on the OS and GPU, so there’s no one-size-fits-all solution. For example, on Windows, there are a few methods that don’t work outside of WSL2…
\nSince the model is FLUX for this project, I recommend the ParaAttention-based optimization mentioned earlier. That alone can significantly speed things up even with a single GPU.
\nAdditionally, combining TorchAO with torch.compile can also improve performance. TorchAO is PyTorch’s official quantization method, so it’s generally fast. However, it’s still a bit unstable in terms of behavior, and selecting the right quantization method requires some knowledge, so it may require some trial and error.
import torch\nfrom diffusers import FluxPipeline, DiffusionPipeline\nimport time, os\nfrom diffusers.quantizers import PipelineQuantizationConfig\nfrom datetime import datetime\n\nstart = time.time()\n\ntorch._dynamo.config.capture_dynamic_output_shape_ops = True\n\n# quantize\npipeline_quant_config = PipelineQuantizationConfig(\n quant_backend=""bitsandbytes_4bit"",\n quant_kwargs={""load_in_4bit"": True, ""bnb_4bit_quant_type"": ""nf4"", ""bnb_4bit_compute_dtype"": torch.bfloat16},\n components_to_quantize=[""transformer"", ""text_encoder_2""],\n)\npipeline = DiffusionPipeline.from_pretrained(\n ""black-forest-labs/FLUX.1-dev"",\n quantization_config=pipeline_quant_config,\n torch_dtype=torch.bfloat16,\n).to(""cuda"")\n\n# compile\npipeline.transformer.to(memory_format=torch.channels_last)\npipeline.enable_model_cpu_offload() # more memory efficient way\n#pipeline.transformer.compile_repeated_blocks(fullgraph=True, dynamic=True) # if you want to compile it\n\nprompt = ""a wolf running"" \n\nimages_ = pipeline(\n prompt,\n width=1920,\n height=1088,\n # width=64,\n # height=64,\n guidance_scale=3.5,\n num_inference_steps=50,\n max_sequence_length=512,\n generator=torch.Generator(device=""cuda"").manual_seed(0)).images\nGitHub - sayakpaul/diffusers-torchao: End-to-end recipes for optimizing diffusion models with torchao and diffusers (inference and FP8 training). (The method you are using for quantization is the new specification for Diffusers, but this document can be useful as a reference for benchmarking and other considerations)
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-21T23:40:55.036Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 163940, 'topic_slug': 'how-long-does-image-generation-with-black-forest-labs-flux-1-dev-take', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/sayakpaul/diffusers-torchao', 'internal': False, 'reflection': False, 'title': 'GitHub - sayakpaul/diffusers-torchao: End-to-end recipes for optimizing diffusion models with torchao and diffusers (inference and FP8 training).', 'clicks': 0}, {'url': 'https://huggingface.co/docs/diffusers/v0.34.0/en/optimization/fp16', 'internal': False, 'reflection': False, 'title': 'Accelerate inference', 'clicks': 0}, {'url': 'https://huggingface.co/docs/diffusers/v0.34.0/en/optimization/speed-memory-optims?offloading=model%2BCPU%2Boffloading', 'internal': False, 'reflection': False, 'title': 'Compile and offloading quantized models', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-long-does-image-generation-with-black-forest-labs-flux-1-dev-take/163940/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234359, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-22T11:40:53.070Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-07-22T11:40:53.070Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 163940, 'topic_slug': 'how-long-does-image-generation-with-black-forest-labs-flux-1-dev-take', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-long-does-image-generation-with-black-forest-labs-flux-1-dev-take/163940/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I run below code on a RTX 3090 with Ryzen 9 7900X and 128 GB RAM. So generating a single 512x512 image takes 20 minutes.
+Is that normal? I read that it just should take seconds.
import torch
+from diffusers import FluxPipeline
+import sys
+import time
+
+start = time.time()
+print(""CUDA available:"", torch.cuda.is_available())
+print(""Device:"", torch.cuda.get_device_name(0) if torch.cuda.is_available() else ""CPU"")
+
+pipe = FluxPipeline.from_pretrained(""black-forest-labs/FLUX.1-dev"", torch_dtype=torch.bfloat16)
+pipe.to(""cuda"")
+
+prompt = ""a wolf running""
+
+images_ = pipe(
+ prompt,
+ # width=1920,
+ # height=1088,
+ width=512,
+ height=512,
+ guidance_scale=3.5,
+ num_inference_steps=50,
+ max_sequence_length=512,
+ generator=torch.Generator(device=""cuda"").manual_seed(0)
+).images
+
+for i, image in enumerate(images_):
+ image.save(""flux-dev"" + str(i) + "".png"")
+
+end = time.time()
+print(f""Generation took {time.time() - start:.2f} seconds"")
+Cuda is 12.1, PYthon is 3.10
+Packages (installed version | lastest version):
| GitPython | +3.1.44 | +3.1.44 | +
|---|---|---|
| MarkupSafe | +2.1.5 | +3.0.2 | +
| PyYAML | +6.0.2 | +6.0.2 | +
| accelerate | +1.9.0 | +1.9.0 | +
| aiofiles | +23.2.1 | +24.1.0 | +
| altair | +5.5.0 | +5.5.0 | +
| annotated-types | +0.7.0 | +0.7.0 | +
| anyio | +4.9.0 | +4.9.0 | +
| attrs | +25.3.0 | +25.3.0 | +
| blinker | +1.9.0 | +1.9.0 | +
| cachetools | +6.1.0 | +6.1.0 | +
| certifi | +2025.7.14 | +2025.7.14 | +
| charset-normalizer | +3.4.2 | +3.4.2 | +
| click | +8.2.1 | +8.2.1 | +
| colorama | +0.4.6 | +0.4.6 | +
| diffusers | +0.34.0 | +0.34.0 | +
| einops | +0.8.1 | +0.8.1 | +
| exceptiongroup | +1.3.0 | +1.3.0 | +
| fastapi | +0.116.1 | +0.116.1 | +
| ffmpy | +0.6.0 | +0.6.0 | +
| filelock | +3.18.0 | +3.18.0 | +
| fire | +0.7.0 | +0.7.0 | +
| flux | +0.0.post58+g1371b2b | +1.3.5 | +
| fsspec | +2025.7.0 | +2025.7.0 | +
| gitdb | +4.0.12 | +4.0.12 | +
| gradio | +5.13.2 | +5.38.0 | +
| gradio-client | +1.6.0 | +1.11.0 | +
| h11 | +0.16.0 | +0.16.0 | +
| httpcore | +1.0.9 | +1.0.9 | +
| httpx | +0.28.1 | +0.28.1 | +
| huggingface-hub | +0.33.4 | +0.33.4 | +
| idna | +3.10 | +3.10 | +
| importlib-metadata | +8.7.0 | +8.7.0 | +
| invisible-watermark | +0.2.0 | +0.2.0 | +
| jinja2 | +3.1.6 | +3.1.6 | +
| jsonschema | +4.25.0 | +4.25.0 | +
| jsonschema-specifications | +2025.4.1 | +2025.4.1 | +
| markdown-it-py | +3.0.0 | +3.0.0 | +
| mdurl | +0.1.2 | +0.1.2 | +
| mpmath | +1.3.0 | +1.3.0 | +
| narwhals | +1.48.0 | +1.48.0 | +
| networkx | +3.4.2 | +3.5 | +
| numpy | +2.2.6 | +2.3.1 | +
| opencv-python | +4.12.0.88 | +4.12.0.88 | +
| orjson | +3.11.0 | +3.11.0 | +
| packaging | +25.0 | +25.0 | +
| pandas | +2.3.1 | +2.3.1 | +
| pillow | +11.3.0 | +11.3.0 | +
| pip | +25.1.1 | +25.1.1 | +
| protobuf | +6.31.1 | +6.31.1 | +
| psutil | +7.0.0 | +7.0.0 | +
| pyarrow | +21.0.0 | +21.0.0 | +
| pydantic | +2.11.7 | +2.11.7 | +
| pydantic-core | +2.33.2 | ++ |
| pydeck | +0.9.1 | +0.9.1 | +
| pydub | +0.25.1 | +0.25.1 | +
| pygments | +2.19.2 | +2.19.2 | +
| python-dateutil | +2.9.0.post0 | +2.9.0.post0 | +
| python-multipart | +0.0.20 | +0.0.20 | +
| pytz | +2025.2 | +2025.2 | +
| pywavelets | +1.8.0 | +1.8.0 | +
| referencing | +0.36.2 | +0.36.2 | +
| regex | +2024.11.6 | +2024.11.6 | +
| requests | +2.32.4 | +2.32.4 | +
| rich | +14.0.0 | +14.0.0 | +
| rpds-py | +0.26.0 | +0.26.0 | +
| ruff | +0.6.8 | +0.12.4 | +
| safehttpx | +0.1.6 | +0.1.6 | +
| safetensors | +0.5.3 | +0.5.3 | +
| semantic-version | +2.10.0 | +2.10.0 | +
| sentencepiece | +0.2.0 | +0.2.0 | +
| setuptools | +57.4.0 | +80.9.0 | +
| shellingham | +1.5.4 | +1.5.4 | +
| six | +1.17.0 | +1.17.0 | +
| smmap | +5.0.2 | +6.0.0 | +
| sniffio | +1.3.1 | +1.3.1 | +
| starlette | +0.47.2 | +0.47.2 | +
| streamlit | +1.47.0 | +1.47.0 | +
| streamlit-drawable-canvas | +0.9.3 | +0.9.3 | +
| streamlit-keyup | +0.3.0 | +0.3.0 | +
| sympy | +1.13.1 | +1.14.0 | +
| tenacity | +9.1.2 | +9.1.2 | +
| termcolor | +3.1.0 | +3.1.0 | +
| tokenizers | +0.21.2 | +0.21.2 | +
| toml | +0.10.2 | +0.10.2 | +
| tomlkit | +0.13.3 | +0.13.3 | +
| torch | +2.5.1+cu121 | +2.7.1 | +
| torchaudio | +2.5.1+cu121 | +2.7.1 | +
| torchvision | +0.20.1+cu121 | +0.22.1 | +
| tornado | +6.5.1 | +6.5.1 | +
| tqdm | +4.67.1 | +4.67.1 | +
| transformers | +4.53.2 | +4.53.2 | +
| typer | +0.16.0 | +0.16.0 | +
| typing-extensions | +4.14.1 | +4.14.1 | +
| typing-inspection | +0.4.1 | +0.4.1 | +
| tzdata | +2025.2 | +2025.2 | +
| urllib3 | +2.5.0 | +2.5.0 | +
| uvicorn | +0.35.0 | +0.35.0 | +
| watchdog | +6.0.0 | +6.0.0 | +
| websockets | +14.2 | +15.0.1 | +
| zipp | +3.23.0 | +3.23.0 | +
++on a RTX 3090 with Ryzen 9 7900X and 128 GB RAM. So generating a single 512x512 image takes 20 minutes.
+
+Is that normal?
Yeah. With that code, FLUX is loaded into VRAM or RAM in a 16-bit state without quantization, requiring approximately 36 GB or more. Since VRAM is insufficient, it cannot be utilized effectively, resulting in lengthy inference times. Therefore,
+Quantization is at least necessary. For 4-bit quantization methods, I recommend BitsAndBytes for ease of use or TorchAO for speed.
+While there were various limitations when using LoRA in the past, these should be largely resolved now.
Optimization methods for FLUX:
+ + +" +Open port for space to connect to PostgreSQL,https://discuss.huggingface.co/t/open-port-for-space-to-connect-to-postgresql/29938,29938,24,2023-01-18 09:09:42.252000+00:00,"[{'id': 55116, 'name': None, 'username': 'anon86412018', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/a698b9/{size}.png', 'created_at': '2023-01-18T09:09:42.333Z', 'cooked': 'Hi @chris-rannou,
\nCould you open the port 5432 for this space: Defi Ai 2022 - a Hugging Face Space by vnghia as I need to connect to a PostgreSQL database ?
Thank you very much !
', 'post_number': 1, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-18T09:09:42.333Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1239, 'reads': 67, 'readers_count': 66, 'score': 6193.4, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': None, 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/vnghia/defi-ai-2022', 'internal': False, 'reflection': False, 'title': 'Defi Ai 2022 - a Hugging Face Space by vnghia', 'clicks': 47}, {'url': 'https://discuss.huggingface.co/t/open-port-9243-on-spaces-to-connect-to-elasticsearch/38699', 'internal': True, 'reflection': True, 'title': 'Open Port 9243 on Spaces to Connect to ElasticSearch', 'clicks': 1}, {'url': 'https://discuss.huggingface.co/t/gprc-on-spaces/152803/3', 'internal': True, 'reflection': True, 'title': 'gPRC on Spaces 🥹', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/problem-summary-hugging-face-space-running-but-line-webhook-verification-fails-with-no-logs/158468/2', 'internal': True, 'reflection': True, 'title': 'Problem Summary: Hugging Face Space Running, but Line Webhook Verification Fails with No Logs', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 14210, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 55140, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-01-18T15:56:29.757Z', 'cooked': 'hi @anon86412018 are you sure your DB service is running at 34.155.175.170:5432? if you’re trying to access the DB from space, you don’t need that port to be open, however on your Space log it states timeout trying to reach your db server
Hi @radames, I am quite sure my DB service is running at 34.155.175.170:5432 because the same code works on my machine. It is a Google Cloud SQL instance (I already opened the DB to every IP and port by 0.0.0.0/0 on GCP side), maybe that is the reason why I have this error ?
ok you’re right, you might need outgoing port access, currently only 80 and 443, we’ll get back to you soon.
', 'post_number': 4, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-18T19:29:57.267Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 60, 'readers_count': 59, 'score': 32.0, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 14210, 'username': 'anon86412018', 'name': None, 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/a698b9/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 55227, 'name': 'Christophe Rannou', 'username': 'chris-rannou', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/7feea3/{size}.png', 'created_at': '2023-01-19T15:42:29.545Z', 'cooked': 'Hi @anon86412018,
\nPort 5432 is now open.
', 'post_number': 5, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-19T15:42:29.545Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 58, 'readers_count': 57, 'score': 61.6, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Christophe Rannou', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 6211, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/5', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 55241, 'name': None, 'username': 'anon86412018', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/a698b9/{size}.png', 'created_at': '2023-01-19T19:13:27.400Z', 'cooked': 'hmmm, unfortuntately, I still can not access to my DB instance. I also add a command to check if the DB is ready by pg_isready. And I found that when building the image, the connection is fine, but it failed while the space is running.
You can see the log here: Defi Ai 2022 - a Hugging Face Space by vnghia
\nDo the port need to be opened twice for building and running or there is something else ?
', 'post_number': 6, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-19T19:13:27.400Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 56, 'readers_count': 55, 'score': 21.2, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': None, 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/vnghia/defi-ai-2022?logs=build', 'internal': False, 'reflection': False, 'title': 'Defi Ai 2022 - a Hugging Face Space by vnghia', 'clicks': 11}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 14210, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 55259, 'name': 'Hyoung-Kyu Song', 'username': 'deepkyu', 'avatar_template': '/user_avatar/discuss.huggingface.co/deepkyu/{size}/19615_2.png', 'created_at': '2023-01-20T04:56:13.139Z', 'cooked': 'Hi @anon86412018 ,
\nI had a similar issue when integrating my Hugging Face Space with my AWS instance.
\nI later found that Hugging Face Space only approves for the privileged port, which is below 1024.
\nI think this is for security reason, and I suggest that you change your SQL server port open with privileged port.
For now, I switched the service port to 80, but I remembered that it is fine if the port number is below 1024.
\nRef for my previous issue:
\n', 'post_number': 7, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-20T04:57:23.852Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 20, 'reads': 51, 'readers_count': 50, 'score': 110.2, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Hyoung-Kyu Song', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/is-there-a-way-to-call-external-grpc-service/14468', 'internal': True, 'reflection': False, 'title': 'Is there a way to call external gRPC service?', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8000, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 55283, 'name': None, 'username': 'anon86412018', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/a698b9/{size}.png', 'created_at': '2023-01-20T10:49:14.149Z', 'cooked': 'Hi @deepkyu I dont think so because @chris-rannou has already opened the port and my code can connect to the database while building the Docker image but not while running. I am suspecting there are some bugs with the Docker space
@anon86412018
\nOh I see. that’s also one of weird situations…
From my experience, I concluded that there were some outbound policies in Hugging Face Space server which blocks unprivileged ports. At that time, my docker container at my AWS instance communicates well from other servers’ request except the HF Space.
\nI’m sorry for not being helpful tho.
\nHope it works out
hi @anon86412018 and @deepkyu , we’ve changed the rules and we’ll enable 5432, 27017 in addition to 80, 443. Sorry @anon86412018 I don’t think it’s in prod yet. I’ll ping you here. Thanks
', 'post_number': 10, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-20T14:24:12.742Z', 'reply_count': 1, 'reply_to_post_number': 9, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 40, 'readers_count': 39, 'score': 63.0, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/open-5432-port-to-connect-to-postgresql-for-langfuse-app/149230/2', 'internal': True, 'reflection': True, 'title': 'Open 5432 port to connect to PostgreSQL for langfuse app', 'clicks': 1}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 8000, 'username': 'deepkyu', 'name': 'Hyoung-Kyu Song', 'avatar_template': '/user_avatar/discuss.huggingface.co/deepkyu/{size}/19615_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/10', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 55313, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-01-20T18:10:02.058Z', 'cooked': 'hi @anon86412018 it should be fixed now, thanks for the patience
', 'post_number': 11, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-20T18:10:02.058Z', 'reply_count': 0, 'reply_to_post_number': 10, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 35, 'readers_count': 34, 'score': 37.0, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/11', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 55315, 'name': None, 'username': 'anon86412018', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/a698b9/{size}.png', 'created_at': '2023-01-20T18:25:31.779Z', 'cooked': 'Thank you very much !
', 'post_number': 12, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-20T18:25:31.779Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 34, 'readers_count': 33, 'score': 21.8, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': None, 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 14210, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/12', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 67686, 'name': 'Karim Foda', 'username': 'kmfoda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png', 'created_at': '2023-05-03T10:21:11.201Z', 'cooked': 'Hey @radames thanks for opening up 5432. I’m hoping to use ElasticSearch (9243) and Papertrail logging (45454) for my app. Would it be possible to open up those 2 ports as well in addition to 5432?
the ports 5432, 9200 and 45454 are now open
', 'post_number': 14, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-05-04T16:54:20.585Z', 'reply_count': 1, 'reply_to_post_number': 13, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 28, 'readers_count': 27, 'score': 15.6, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 298, 'username': 'kmfoda', 'name': 'Karim Foda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/14', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 67929, 'name': 'Karim Foda', 'username': 'kmfoda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png', 'created_at': '2023-05-04T16:55:38.679Z', 'cooked': 'Sorry my apologies I mean 9243 not 9200. I believe that’s the port Elastic uses. Thanks so much!
', 'post_number': 15, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-05-04T16:55:38.679Z', 'reply_count': 1, 'reply_to_post_number': 14, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 29, 'readers_count': 28, 'score': 15.8, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Karim Foda', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 298, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/15', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 67930, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-05-04T16:57:24.180Z', 'cooked': 'I see, I guess the default ES port is 9200 and it’s been open already, could you change it on your app?
', 'post_number': 16, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-05-04T16:57:24.180Z', 'reply_count': 1, 'reply_to_post_number': 15, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 30, 'readers_count': 29, 'score': 21.0, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 298, 'username': 'kmfoda', 'name': 'Karim Foda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/16', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 67934, 'name': 'Karim Foda', 'username': 'kmfoda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png', 'created_at': '2023-05-04T17:34:34.265Z', 'cooked': 'Ah we’re running our app on elastic.co and that’s the port they gave us unfortunately. I think it might be quite tricky for us to change the port, it’ll also have a bit of downstream impact on all our other services which we’d have to factor in.
', 'post_number': 17, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-05-04T17:34:34.265Z', 'reply_count': 1, 'reply_to_post_number': 16, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 29, 'readers_count': 28, 'score': 30.8, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Karim Foda', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://elastic.co/', 'internal': False, 'reflection': False, 'title': 'Elastic Observability and Security — built on Elasticsearch | Elastic', 'clicks': 11}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 298, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/17', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 68064, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-05-05T16:54:03.492Z', 'cooked': 'hi @kmfoda , the requested ports are open now, please try it again. Thanks
', 'post_number': 18, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-05-05T16:54:03.492Z', 'reply_count': 0, 'reply_to_post_number': 17, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 27, 'readers_count': 26, 'score': 10.4, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 298, 'username': 'kmfoda', 'name': 'Karim Foda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/18', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 68070, 'name': 'Karim Foda', 'username': 'kmfoda', 'avatar_template': '/user_avatar/discuss.huggingface.co/kmfoda/{size}/42122_2.png', 'created_at': '2023-05-05T18:01:45.239Z', 'cooked': 'Hi @radames, amazing that worked now! Thank you very much for your help!
', 'post_number': 19, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-05-05T18:01:45.239Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 27, 'readers_count': 26, 'score': 40.4, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Karim Foda', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 298, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/19', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234263, 'name': 'Notionhive AI', 'username': 'notionhive-ai', 'avatar_template': '/user_avatar/discuss.huggingface.co/notionhive-ai/{size}/51497_2.png', 'created_at': '2025-07-22T06:51:20.965Z', 'cooked': 'Hi @radames, is there any way to open the port 587 for mail SMTP and 443 port to communicate through telegram?
', 'post_number': 20, 'post_type': 1, 'posts_count': 20, 'updated_at': '2025-07-22T06:51:20.965Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 29938, 'topic_slug': 'open-port-for-space-to-connect-to-postgresql', 'display_username': 'Notionhive AI', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99997, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/open-port-for-space-to-connect-to-postgresql/29938/20', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi @chris-rannou,
+Could you open the port 5432 for this space: Defi Ai 2022 - a Hugging Face Space by vnghia as I need to connect to a PostgreSQL database ?
Thank you very much !
","hi @anon86412018 it should be fixed now, thanks for the patience
" +Recommendations for ML courses,https://discuss.huggingface.co/t/recommendations-for-ml-courses/163811,163811,5,2025-07-20 11:40:24.641000+00:00,"[{'id': 233967, 'name': 'Anisimov', 'username': 'kaguya3222', 'avatar_template': '/user_avatar/discuss.huggingface.co/kaguya3222/{size}/51401_2.png', 'created_at': '2025-07-20T11:40:24.705Z', 'cooked': 'Hey there ! I am Maksym, Frontend Engineer. I have 5 years of experience and working mostly with TypeScript/Frontend frameworks. I am familiar with other languages (C, C++) from the university program. I am interested in learning basic ML to complete Hugging Face LLM Course.
\nAny recommendations here with what should I start?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-20T11:40:24.705Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 9, 'readers_count': 8, 'score': 101.8, 'yours': False, 'topic_id': 163811, 'topic_slug': 'recommendations-for-ml-courses', 'display_username': 'Anisimov', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99851, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/recommendations-for-ml-courses/163811/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 233983, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-20T14:05:26.387Z', 'cooked': 'Hi.
\nPython is essential. However, you don’t necessarily need to study Python specifically; if you understand C, you should be able to use Python by looking up syntax and functions as needed. The course should not use many of the more complex Python syntaxes. (After all, Python’s strengths lie in its readability and abundance of libraries…)
\nYou can start right away without any issues.
Additionally, for actual API usage or running WebGPU in a browser, there are JavaScript libraries available.
\nIf you want to learn the theoretical background, there are other resources available, but the LLM course alone covers a significant portion of the material.
\nThanks a lot!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-20T14:24:42.104Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 163811, 'topic_slug': 'recommendations-for-ml-courses', 'display_username': 'Anisimov', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99851, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/recommendations-for-ml-courses/163811/3', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 234048, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-21T02:25:23.946Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-21T02:25:23.946Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 163811, 'topic_slug': 'recommendations-for-ml-courses', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/recommendations-for-ml-courses/163811/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hey there ! I am Maksym, Frontend Engineer. I have 5 years of experience and working mostly with TypeScript/Frontend frameworks. I am familiar with other languages (C, C++) from the university program. I am interested in learning basic ML to complete Hugging Face LLM Course.
+Any recommendations here with what should I start?
","Hi.
+Python is essential. However, you don’t necessarily need to study Python specifically; if you understand C, you should be able to use Python by looking up syntax and functions as needed. The course should not use many of the more complex Python syntaxes. (After all, Python’s strengths lie in its readability and abundance of libraries…)
+You can start right away without any issues.
Additionally, for actual API usage or running WebGPU in a browser, there are JavaScript libraries available.
+If you want to learn the theoretical background, there are other resources available, but the LLM course alone covers a significant portion of the material.
+I have been asking a few AI on how to do it, seems like the code they provided would give execution errors.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-19T13:21:14.185Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 15, 'reads': 7, 'readers_count': 6, 'score': 81.4, 'yours': False, 'topic_id': 163714, 'topic_slug': 'are-there-any-recommendation-tutorials-on-how-to-train-a-llm-via-colab', 'display_username': 'bun', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99788, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/are-there-any-recommendation-tutorials-on-how-to-train-a-llm-via-colab/163714/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 233850, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-19T13:53:53.109Z', 'cooked': 'I recommend trying the LLM course. It basically uses Colab. Of course, if you have a good GPU, you can do it locally…
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-19T13:53:53.109Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 163714, 'topic_slug': 'are-there-any-recommendation-tutorials-on-how-to-train-a-llm-via-colab', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/learn/llm-course/en/chapter3/3', 'internal': False, 'reflection': False, 'title': 'Fine-tuning a model with the Trainer API - Hugging Face LLM Course', 'clicks': 3}, {'url': 'https://huggingface.co/blog/dvgodoy/fine-tuning-llm-hugging-face', 'internal': False, 'reflection': False, 'title': 'Fine-Tuning Your First Large Language Model (LLM) with PyTorch and Hugging Face', 'clicks': 1}, {'url': 'https://huggingface.co/docs/transformers/en/notebooks', 'internal': False, 'reflection': False, 'title': '🤗 Transformers Notebooks', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/are-there-any-recommendation-tutorials-on-how-to-train-a-llm-via-colab/163714/2', 'reactions': [{'id': 'white_check_mark', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 233923, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-20T04:01:51.141Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-20T04:01:51.141Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 163714, 'topic_slug': 'are-there-any-recommendation-tutorials-on-how-to-train-a-llm-via-colab', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/are-there-any-recommendation-tutorials-on-how-to-train-a-llm-via-colab/163714/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have been asking a few AI on how to do it, seems like the code they provided would give execution errors.
","I recommend trying the LLM course. It basically uses Colab. Of course, if you have a good GPU, you can do it locally…
+ + +" +Inconsistent GPT2Model results between transformers versions,https://discuss.huggingface.co/t/inconsistent-gpt2model-results-between-transformers-versions/163484,163484,6,2025-07-17 16:01:05.497000+00:00,"[{'id': 233493, 'name': 'Wenzhong Zhao', 'username': 'Wenzhong2005', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/w/b3f665/{size}.png', 'created_at': '2025-07-17T16:01:05.596Z', 'cooked': 'We fine-tuned the GPT2Model (distilgpt2) some time ago. The exact same GPT2 model produces different outputs for the exact same input after the upgrading. Therefore, after applying a classification head (linear layer) on top of GPT-2 output, we got different scores for the same input. It seems to me that the masked portion of the model output changed, while the unmasked portion stays the same. In the past upgrade, we have seen the default value for the attn_implementation changed from “eager” to “sdpa”. See my previous topic. Due to tool vulnerability issues, we have to upgrade transformers 4.52.3 or above. This time, I already specified attn_implementation=“eager”, I still got different results after the upgrade. Can anyone help to point to what’s changed?
\nThe code to reproduce the results:
\nimport torch
\nimport tokenizers
\nimport transformers
\nfrom transformers import GPT2Model, GPT2Tokenizer
#Sample input
\ntokenizer = GPT2Tokenizer.from_pretrained(‘distilgpt2’)
\ntokenizer.pad_token = tokenizer.eos_token
\ntokenizer.padding_side = ‘left’
text = ‘DAVID DAVIS’
\nmodel_inputs = tokenizer(text, padding=‘max_length’, max_length=12, truncation=True, return_tensors=‘pt’)
\ninput_ids, attention_mask = model_inputs[‘input_ids’],model_inputs[‘attention_mask’]
\nprint(‘input_ids:’, input_ids)
\nprint(‘mask:’, attention_mask)
#Load GPT-2 Model
\nmodel = GPT2Model.from_pretrained(‘distilgpt2’, attn_implementation=“eager”)
#Run model
\nmodel.eval()
\nwith torch.no_grad():
\noutputs = model(input_ids=input_ids, attention_mask=attention_mask)
last_hidden_state = outputs.last_hidden_state
\nprint(last_hidden_state)
Here are the 2 requirements.txt files and model outputs:
\nBefore:
\ntorch==2.6.0
\ntransformers==4.50.0
\nhuggingface_hub==0.33.4
input_ids: tensor([[50256, 50256, 50256, 50256, 50256, 50256, 50256, 50256, 5631, 11008, 42274, 1797]])
\nmask: tensor([[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1]])
\nModel output: tensor([[[-3.1153e-01, 1.1569e-01, 2.4667e-02, …, -1.6813e-01, -1.9119e-01, -4.2739e-02],
\n[-8.7119e-01, 2.1186e-04, 5.6834e-01, …, -1.1233e-01, -4.8243e-01, 4.7066e-02],
\n[-7.1241e-01, -4.7743e-02, 5.6767e-01, …, 1.0435e-02, -4.7335e-01, 2.1707e-04],
\n…,
\n[-1.3753e+00, 2.9666e-01, 5.7950e-01, …, -6.4851e-01, -1.2977e+00, -8.4751e-02],
\n[-1.2291e+00, 1.6299e-01, 4.4637e-01, …, -5.1411e-01, -6.0615e-01, 4.3908e-01],
\n[-1.3633e+00, 8.3929e-02, 5.4821e-01, …, -5.7178e-01, -6.4784e-01, 4.6220e-01]]])
After:
\ntorch==2.6.0
\ntransformers==4.52.3
\nhuggingface_hub==0.33.4
input_ids: tensor([[50256, 50256, 50256, 50256, 50256, 50256, 50256, 50256, 5631, 11008, 42274, 1797]])
\nmask: tensor([[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1]])
\nModel output: tensor([[[-0.0724, 0.4212, 0.0130, …, -0.1462, 0.1229, -0.0698],
\n[-0.0360, 0.4466, -0.0973, …, -0.0136, 0.1273, -0.0545],
\n[ 0.0104, 0.3948, -0.0099, …, 0.0273, 0.1091, -0.0364],
\n…,
\n[-1.3753, 0.2967, 0.5795, …, -0.6485, -1.2978, -0.0848],
\n[-1.2291, 0.1630, 0.4464, …, -0.5141, -0.6062, 0.4391],
\n[-1.3633, 0.0839, 0.5482, …, -0.5718, -0.6479, 0.4622]]])
Although not mentioned in the release notes, it appears that the implementation of masks and attention has been significantly changed…
', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-07-18T00:03:07.980Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 6.4, 'yours': False, 'topic_id': 163484, 'topic_slug': 'inconsistent-gpt2model-results-between-transformers-versions', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/commits/main/src/transformers/models/gpt2/modeling_gpt2.py', 'internal': False, 'reflection': False, 'title': 'History for src/transformers/models/gpt2/modeling_gpt2.py - huggingface/transformers · GitHub', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inconsistent-gpt2model-results-between-transformers-versions/163484/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233563, 'name': 'Wenzhong Zhao', 'username': 'Wenzhong2005', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/w/b3f665/{size}.png', 'created_at': '2025-07-18T00:30:57.149Z', 'cooked': '@John6666 thanks for the response. I figured that the latest version has the correct implementation for masks and attention: both from padded to non-padded tokens and other way around. I think we better to use the latest version to rebuild the fine-tuned model in the long term. However, for security reasons we need to upgrade it now, and the performance impact is too big to be ignored. Are there any workaround on this issue?
', 'post_number': 3, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-07-18T00:43:10.026Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 16.6, 'yours': False, 'topic_id': 163484, 'topic_slug': 'inconsistent-gpt2model-results-between-transformers-versions', 'display_username': 'Wenzhong Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 22921, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inconsistent-gpt2model-results-between-transformers-versions/163484/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233574, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-18T03:03:36.358Z', 'cooked': 'Since we can get the same output by using the same code, there are two options: simply download the old version of the source code and replace it, or fork Transformers and revert only the specific changes.
\nAnother option is a monkey patch like the one below. I haven’t confirmed whether it works or not…
\n# full_monkey_patch_gpt2_mask.py\n\nimport torch\nfrom transformers import GPT2Model, GPT2Tokenizer\nfrom transformers.modeling_attn_mask_utils import AttentionMaskConverter\n\n# ─── 1. Legacy v4.50.0 mask helpers ───────────────────────────────────────────\n# Copied from https://raw.githubusercontent.com/huggingface/transformers/v4.50.0/.../modeling_attn_mask_utils.py\n\ndef old_expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: int = None):\n bsz, src_len = mask.size()\n tgt_len = tgt_len if tgt_len is not None else src_len\n expanded = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype)\n inv = 1.0 - expanded\n return inv.masked_fill(inv.to(torch.bool), torch.finfo(dtype).min)\n\ndef old_to_causal_4d(\n attention_mask: torch.Tensor,\n input_shape: tuple[int, int],\n inputs_embeds: torch.Tensor,\n past_key_values_length: int,\n sliding_window: int | None = None,\n):\n # Reconstruct converter usage from v4.50.0\n converter = AttentionMaskConverter(is_causal=True, sliding_window=sliding_window)\n key_value_length = input_shape[-1] + past_key_values_length\n if attention_mask is not None and attention_mask.dim() == 2:\n return converter.to_4d(\n attention_mask,\n input_shape[-1],\n key_value_length=key_value_length,\n dtype=inputs_embeds.dtype,\n )\n return converter.to_causal_4d(\n input_shape[0],\n input_shape[-1],\n key_value_length,\n dtype=inputs_embeds.dtype,\n device=inputs_embeds.device,\n )\n\n# ─── 2. Monkey-patch the new converter ────────────────────────────────────────\n# This forces Transformers ≥ 4.51 to use our old logic instead of the refactored one\n\nAttentionMaskConverter._expand_mask = staticmethod(old_expand_mask)\nAttentionMaskConverter.to_causal_4d = staticmethod(old_to_causal_4d)\nAttentionMaskConverter.to_4d = staticmethod(lambda mask, qlen, key_value_length=None, dtype=None: \n old_expand_mask(mask, dtype, tgt_len=qlen))\n\n# Prevent SDPA from dropping masks on trivial sequences:\nAttentionMaskConverter._ignore_causal_mask_sdpa = staticmethod(lambda *args, **kwargs: False)\nThanks @John6666. Tried the above monkey patch you provided, but it does not change the model output.
', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-07-18T17:37:08.676Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 163484, 'topic_slug': 'inconsistent-gpt2model-results-between-transformers-versions', 'display_username': 'Wenzhong Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 22921, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inconsistent-gpt2model-results-between-transformers-versions/163484/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233758, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-18T23:47:31.304Z', 'cooked': 'As a last resort, downloading this file and saving it locally should allow you to import the old version of GPT2Model. Compared to forking and reversing committing, this method is slightly less consistent, but it has the advantage of not being affected by version updates.
\nThe import statements at the beginning can be rewritten to suit your environment.
Additionally, you could simply copy and paste the code from the old version, define the GPT2Model class, and use it. Since the modules are designed to have minimal dependencies on each other, the implementation should not be too difficult.
\nIf we decide to use AutoModel, there will be an extra step, but if we only use GPT2Model, defining the class is all that’s needed.
Thanks @John6666 This is a good recommendation. We had a workaround with a slightly lower version v4.51.3 which still satisfies our security requirements. So it is fine for now.
', 'post_number': 7, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-07-19T03:25:05.274Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 163484, 'topic_slug': 'inconsistent-gpt2model-results-between-transformers-versions', 'display_username': 'Wenzhong Zhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 22921, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inconsistent-gpt2model-results-between-transformers-versions/163484/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233861, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-19T15:26:01.130Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-07-19T15:26:01.130Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 163484, 'topic_slug': 'inconsistent-gpt2model-results-between-transformers-versions', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/inconsistent-gpt2model-results-between-transformers-versions/163484/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","We fine-tuned the GPT2Model (distilgpt2) some time ago. The exact same GPT2 model produces different outputs for the exact same input after the upgrading. Therefore, after applying a classification head (linear layer) on top of GPT-2 output, we got different scores for the same input. It seems to me that the masked portion of the model output changed, while the unmasked portion stays the same. In the past upgrade, we have seen the default value for the attn_implementation changed from “eager” to “sdpa”. See my previous topic. Due to tool vulnerability issues, we have to upgrade transformers 4.52.3 or above. This time, I already specified attn_implementation=“eager”, I still got different results after the upgrade. Can anyone help to point to what’s changed?
+The code to reproduce the results:
+import torch
+import tokenizers
+import transformers
+from transformers import GPT2Model, GPT2Tokenizer
#Sample input
+tokenizer = GPT2Tokenizer.from_pretrained(‘distilgpt2’)
+tokenizer.pad_token = tokenizer.eos_token
+tokenizer.padding_side = ‘left’
text = ‘DAVID DAVIS’
+model_inputs = tokenizer(text, padding=‘max_length’, max_length=12, truncation=True, return_tensors=‘pt’)
+input_ids, attention_mask = model_inputs[‘input_ids’],model_inputs[‘attention_mask’]
+print(‘input_ids:’, input_ids)
+print(‘mask:’, attention_mask)
#Load GPT-2 Model
+model = GPT2Model.from_pretrained(‘distilgpt2’, attn_implementation=“eager”)
#Run model
+model.eval()
+with torch.no_grad():
+outputs = model(input_ids=input_ids, attention_mask=attention_mask)
last_hidden_state = outputs.last_hidden_state
+print(last_hidden_state)
Here are the 2 requirements.txt files and model outputs:
+Before:
+torch==2.6.0
+transformers==4.50.0
+huggingface_hub==0.33.4
input_ids: tensor([[50256, 50256, 50256, 50256, 50256, 50256, 50256, 50256, 5631, 11008, 42274, 1797]])
+mask: tensor([[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1]])
+Model output: tensor([[[-3.1153e-01, 1.1569e-01, 2.4667e-02, …, -1.6813e-01, -1.9119e-01, -4.2739e-02],
+[-8.7119e-01, 2.1186e-04, 5.6834e-01, …, -1.1233e-01, -4.8243e-01, 4.7066e-02],
+[-7.1241e-01, -4.7743e-02, 5.6767e-01, …, 1.0435e-02, -4.7335e-01, 2.1707e-04],
+…,
+[-1.3753e+00, 2.9666e-01, 5.7950e-01, …, -6.4851e-01, -1.2977e+00, -8.4751e-02],
+[-1.2291e+00, 1.6299e-01, 4.4637e-01, …, -5.1411e-01, -6.0615e-01, 4.3908e-01],
+[-1.3633e+00, 8.3929e-02, 5.4821e-01, …, -5.7178e-01, -6.4784e-01, 4.6220e-01]]])
After:
+torch==2.6.0
+transformers==4.52.3
+huggingface_hub==0.33.4
input_ids: tensor([[50256, 50256, 50256, 50256, 50256, 50256, 50256, 50256, 5631, 11008, 42274, 1797]])
+mask: tensor([[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1]])
+Model output: tensor([[[-0.0724, 0.4212, 0.0130, …, -0.1462, 0.1229, -0.0698],
+[-0.0360, 0.4466, -0.0973, …, -0.0136, 0.1273, -0.0545],
+[ 0.0104, 0.3948, -0.0099, …, 0.0273, 0.1091, -0.0364],
+…,
+[-1.3753, 0.2967, 0.5795, …, -0.6485, -1.2978, -0.0848],
+[-1.2291, 0.1630, 0.4464, …, -0.5141, -0.6062, 0.4391],
+[-1.3633, 0.0839, 0.5482, …, -0.5718, -0.6479, 0.4622]]])
As a last resort, downloading this file and saving it locally should allow you to import the old version of GPT2Model. Compared to forking and reversing committing, this method is slightly less consistent, but it has the advantage of not being affected by version updates.
+The import statements at the beginning can be rewritten to suit your environment.
Additionally, you could simply copy and paste the code from the old version, define the GPT2Model class, and use it. Since the modules are designed to have minimal dependencies on each other, the implementation should not be too difficult.
+If we decide to use AutoModel, there will be an extra step, but if we only use GPT2Model, defining the class is all that’s needed.
I have developed a method for AI to parse ethics algorithmically.
\nEthics should be open source. I have been developing this in a silo for 12 months, this is my first-ever software project, in the 12 months since I started this journey at “Hello world,” I have not managed to have a meaningful conversation with anyone about this, either from lack of interest, lack of understanding, or hostility because I’m not actually a software developer, and I would genuinely appreciate human feedback on this project, good bad and ugly. Is there an appropriate subforum to post this? Thank you so much!
\n\n', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-17T04:37:54.887Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 10, 'reads': 15, 'readers_count': 14, 'score': 68.0, 'yours': False, 'topic_id': 163372, 'topic_slug': 'i-made-a-thing-and-have-no-idea-what-to-do-now', 'display_username': 'Glen Bradley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/GlenABradley/EthicalAITestbed', 'internal': False, 'reflection': False, 'title': 'GitHub - GlenABradley/EthicalAITestbed: This is Ethics for AI. Not guardrails, actual ethics.', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99577, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-made-a-thing-and-have-no-idea-what-to-do-now/163372/1', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 233429, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-17T13:36:47.294Z', 'cooked': 'Hugging Face Discord has a dedicated channel for AI ethics.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-17T13:36:47.294Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 22.2, 'yours': False, 'topic_id': 163372, 'topic_slug': 'i-made-a-thing-and-have-no-idea-what-to-do-now', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-made-a-thing-and-have-no-idea-what-to-do-now/163372/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 233542, 'name': 'Glen Bradley', 'username': 'glenbradley', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/g/c2a13f/{size}.png', 'created_at': '2025-07-17T21:28:21.212Z', 'cooked': 'Thank you. I am brand new and don’t know my way around yet. I appreciate your help.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-17T21:28:21.212Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 16.8, 'yours': False, 'topic_id': 163372, 'topic_slug': 'i-made-a-thing-and-have-no-idea-what-to-do-now', 'display_username': 'Glen Bradley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99577, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/i-made-a-thing-and-have-no-idea-what-to-do-now/163372/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233644, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-18T09:29:16.259Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-18T09:29:16.259Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 6.2, 'yours': False, 'topic_id': 163372, 'topic_slug': 'i-made-a-thing-and-have-no-idea-what-to-do-now', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/i-made-a-thing-and-have-no-idea-what-to-do-now/163372/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have developed a method for AI to parse ethics algorithmically.
+Ethics should be open source. I have been developing this in a silo for 12 months, this is my first-ever software project, in the 12 months since I started this journey at “Hello world,” I have not managed to have a meaningful conversation with anyone about this, either from lack of interest, lack of understanding, or hostility because I’m not actually a software developer, and I would genuinely appreciate human feedback on this project, good bad and ugly. Is there an appropriate subforum to post this? Thank you so much!
+ +",Hugging Face Discord has a dedicated channel for AI ethics.
+Pipeline vs model.generate(),https://discuss.huggingface.co/t/pipeline-vs-model-generate/26203,26203,5,2022-11-16 22:12:08.333000+00:00,"[{'id': 49588, 'name': 'Zeke John', 'username': 'Z3K3', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/a3d4f5/{size}.png', 'created_at': '2022-11-16T22:12:08.404Z', 'cooked': 'I want to know whats the difference between using the Pipeline() function to generate a result Vs using the model.generate() function to generate a result, which one is faster? Which one is more accurate? Which one is more consistently giving out good responses? And what is the main difference between them. I am sorry if this sounds like a dumb question i am just wondering which method i should use to generate ML predictions for Summarization, and want to know the Pros/Cons of each of them.
\nThanks in advance
', 'post_number': 1, 'post_type': 1, 'posts_count': 12, 'updated_at': '2022-11-16T22:12:08.404Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 14510, 'reads': 448, 'readers_count': 447, 'score': 72499.6, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'Zeke John', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 7}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8150, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/1', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 6}, {'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 7, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 49611, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2022-11-17T08:01:47.700Z', 'cooked': 'Hi,
\nThe pipeline() API is created mostly for people who don’t care too much about the details of the underlying process, for people who just want to use a machine learning model without having to implement several details like pre- and postprocessing themselves. The pipeline API is created such that you get an easy-to-use abstraction over any ML model, which is great for inference. The SummarizationPipeline for instance uses generate() behind the scenes.
\nOn the other hand, if you do care about the details, then it’s recommended to generate text yourself by calling generate() yourself and implement pre-and postprocessing yourself.
\nAlso note that any text generation pipeline does provide a generate_kwargs argument, which means that technically you can forward any of the keyword arguments that generate() supports to the pipeline as well.
', 'post_number': 2, 'post_type': 1, 'posts_count': 12, 'updated_at': '2022-11-17T08:01:47.700Z', 'reply_count': 3, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 272, 'reads': 441, 'readers_count': 440, 'score': 1688.2, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/v4.24.0/en/main_classes/text_generation#transformers.generation_utils.GenerationMixin.generate', 'internal': False, 'reflection': False, 'title': 'Generation', 'clicks': 594}, {'url': 'https://github.com/huggingface/transformers/blob/94b3f544a1f5e04b78d87a2ae32a7ac252e22e31/src/transformers/pipelines/text2text_generation.py#L138', 'internal': False, 'reflection': False, 'title': 'transformers/text2text_generation.py at 94b3f544a1f5e04b78d87a2ae32a7ac252e22e31 · huggingface/transformers · GitHub', 'clicks': 275}, {'url': 'https://huggingface.co/docs/transformers/v4.24.0/en/main_classes/pipelines', 'internal': False, 'reflection': False, 'title': 'Pipelines', 'clicks': 275}, {'url': 'https://huggingface.co/docs/transformers/v4.24.0/en/main_classes/pipelines#transformers.SummarizationPipeline', 'internal': False, 'reflection': False, 'title': 'Pipelines', 'clicks': 130}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 15}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 12}, {'id': '+1', 'type': 'emoji', 'count': 2}, {'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 15, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 49670, 'name': 'Zeke John', 'username': 'Z3K3', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/a3d4f5/{size}.png', 'created_at': '2022-11-17T17:40:09.038Z', 'cooked': 'Thank you for this response nielsr. This was what I wanted to know.
', 'post_number': 3, 'post_type': 1, 'posts_count': 12, 'updated_at': '2022-11-17T17:40:09.038Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 122, 'reads': 419, 'readers_count': 418, 'score': 683.8, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'Zeke John', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8150, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 84585, 'name': 'Saptarshi Sengupta', 'username': 'Saptarshi7', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9e8a1a/{size}.png', 'created_at': '2023-08-16T21:45:20.578Z', 'cooked': 'Hello,
\nSo I tested both recently and found a very peculiar behavior under similar parameter values. This was using Galactica’s 1.3B variant
\nfrom transformers import AutoTokenizer, AutoModelForCausalLM, pipeline, set_seed\nimport torch\n\ncheckpoint = ""facebook/galactica-1.3b""\n\ntokenizer = AutoTokenizer.from_pretrained(checkpoint, padding_side=""left"") \nmodel = AutoModelForCausalLM.from_pretrained(checkpoint)\nmodel.to(\'cuda\')\ngenerator = pipeline(\'text-generation\', model=model, tokenizer=tokenizer, device=0)\n\n#With pipeline\nset_seed(42)\ngenerator([\'Is this\', \'What is the matter\'], renormalize_logits=True, do_sample=True, use_cache=True, max_new_tokens=10)\n\n#With model.generate()\ndevice=torch.device(\'cuda\',0)\nmodel.to(device)\n\ntokenizer = AutoTokenizer.from_pretrained(checkpoint, padding_side=""left"")\ntokenizer.pad_token = tokenizer.eos_token = \'<pad>\'\n\ntokenized_prompts = tokenizer([\'Is this\', \'What is the matter\'], padding=True, return_tensors=\'pt\')\nset_seed(42)\nmodel_op = model.generate(input_ids=tokenized_prompts[\'input_ids\'].to(device),\n attention_mask=tokenized_prompts[\'attention_mask\'].to(device),\n renormalize_logits=False, do_sample=True,\n use_cache=True, max_new_tokens=10)\ntokenizer.batch_decode(model_op, skip_special_tokens=True)\nHere is the result with each,
\n[{\'generated_text\': \'Is this method for dealing with multiple objects?\\n\\n\\n\'}],\n [{\'generated_text\': \'What is the matter density of a star whose radius is equal to \'}]\n................\n[\'Is this method for dealing with multiple objects?\\n\\n\\n\',\n \'What is the matter of this, I know that it isn’t\']\nAs we can see, both methods are producing different outputs, even under the same settings. However, the first generation for each method seems to be the same & I tried it for a bunch of other prompts. That being said if we turn off do_sample i.e.
\n\n\ndo_sample = False (greedy decoding)
\n
then, we get the same results. Thus, I believe this is related to the sampling method being employed which is producing different results. Does anyone have any thoughts on this?
', 'post_number': 4, 'post_type': 1, 'posts_count': 12, 'updated_at': '2023-08-16T21:45:20.578Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 534, 'reads': 351, 'readers_count': 350, 'score': 2775.2, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'Saptarshi Sengupta', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 26605, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'clap', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 105523, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2023-12-25T20:59:13.271Z', 'cooked': 'Hi,
\nWell, sampling is exactly causing randomness you can set a seed to get reproducabile results even when using sampling:
from transformers import set_seed\nset_seed(42)\nRefer to the generate blog post for more details.
', 'post_number': 5, 'post_type': 1, 'posts_count': 12, 'updated_at': '2023-12-25T20:59:13.271Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 94, 'reads': 207, 'readers_count': 206, 'score': 511.4, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/blog/how-to-generate', 'internal': False, 'reflection': False, 'title': 'How to generate text: using different decoding methods for language generation with Transformers', 'clicks': 132}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 26605, 'username': 'Saptarshi7', 'name': 'Saptarshi Sengupta', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/9e8a1a/{size}.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 186805, 'name': 'Brando Miranda', 'username': 'brando', 'avatar_template': '/user_avatar/discuss.huggingface.co/brando/{size}/30114_2.png', 'created_at': '2024-12-05T19:26:49.723Z', 'cooked': '\nDo you mind sharing a concrete example of what you mean by pre and postprocessing in this context? @nielsr
\nThank you in advance.
', 'post_number': 6, 'post_type': 1, 'posts_count': 12, 'updated_at': '2024-12-05T19:26:49.723Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 1, 'incoming_link_count': 15, 'reads': 57, 'readers_count': 56, 'score': 121.4, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'Brando Miranda', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 3664, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/6', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 192327, 'name': 'Niels Rogge', 'username': 'nielsr', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png', 'created_at': '2024-12-29T11:07:37.068Z', 'cooked': 'By pre-processing, I mean turning a sentence into tokens, then turning those tokens into numbers (indices in the vocabulary of a Transformer model). The tokenizer can be used for this purpose, which automatically turns text into so-called input_ids. The pipeline uses a tokenizer behind the scenes.
As for post-processing, one needs to decode the generate id’s back into text. The tokenizer can also be used for this, using the decode or batch_decode methods. The pipeline also makes use of these methods to present the result as text.
Thank you for your response earlier. I have a question regarding the generate_kwargs argument needed to make .generate perform equivalently to .pipeline.
\nCurrently, I am using the model from Meta-Llama-3.1-8B-Instruct-bnb-4bit. When I use .generate, the output begins by repeating the input prompt before generating the desired output. Since my prompt is quite lengthy, I can only see a truncated version of it in the output.
\nHowever, when I use .pipeline, it outputs the desired response directly without repeating the prompt. I suspect the difference might be due to .generate using greedy search for decoding, while .pipeline applies additional configurations like penalty terms to avoid regenerating the prompt.
\nI understand from your response that this might be the case, but I am unsure how to inspect the configuration used by .pipeline and apply similar settings to the model.generation_config. Could you provide an example code snippet illustrating how to achieve this?
\nThank you for your help!
', 'post_number': 8, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-01-20T02:24:33.522Z', 'reply_count': 2, 'reply_to_post_number': 2, 'quote_count': 1, 'incoming_link_count': 15, 'reads': 35, 'readers_count': 34, 'score': 122.0, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'hongyeliu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit', 'internal': False, 'reflection': False, 'title': 'unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit · Hugging Face', 'clicks': 1}, {'url': 'https://github.com/huggingface/transformers/blob/94b3f544a1f5e04b78d87a2ae32a7ac252e22e31/src/transformers/pipelines/text2text_generation.py#L138', 'internal': False, 'reflection': False, 'title': 'transformers/src/transformers/pipelines/text2text_generation.py at 94b3f544a1f5e04b78d87a2ae32a7ac252e22e31 · huggingface/transformers · GitHub', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 67971, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/8', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 203160, 'name': 'hongyeliu', 'username': 'hongyeliu', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/h/ee59a6/{size}.png', 'created_at': '2025-02-17T15:11:48.247Z', 'cooked': '@nielsr sry, forgot to @
', 'post_number': 9, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-02-17T15:11:48.247Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 20, 'readers_count': 19, 'score': 34.0, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'hongyeliu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 67971, 'username': 'hongyeliu', 'name': 'hongyeliu', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/h/ee59a6/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 67971, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231146, 'name': 'bendangnuksung', 'username': 'Bendang', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/a4c791/{size}.png', 'created_at': '2025-07-05T13:50:23.607Z', 'cooked': '\nI am having the same problem. Have you figured out how to do this?
', 'post_number': 10, 'post_type': 1, 'posts_count': 12, 'updated_at': '2025-07-05T13:50:23.607Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 1, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 26.0, 'yours': False, 'topic_id': 26203, 'topic_slug': 'pipeline-vs-model-generate', 'display_username': 'bendangnuksung', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98237, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pipeline-vs-model-generate/26203/10', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231215, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-06T03:55:29.738Z', 'cooked': 'For now, I think the default value in Pipeline is prioritized by generation_config.json, followed by the default value in GenerationConfig. If you reproduce this, you should get almost the same result. Probably like this:
outputs = model.generate(input_ids, do_sample=True, top_k=50, top_p=0.9, temperature=0.6, repetition_penalty=1.0, max_length=131072, bos_token_id=128000, pad_token_id=128004, eos_token_id=[128001, 128008, 128009])\nI found a workaround to make model.generate produce the same output as the pipeline. I ran the pipeline in debug mode and set a breakpoint here. At that point, I pickled the generate_kwargs used internally by the pipeline and reused them directly in my own call to model.generate. This way, I was able to replicate the exact same output as the pipeline.
\nHope this helps anyone facing a similar issue.
I want to know whats the difference between using the Pipeline() function to generate a result Vs using the model.generate() function to generate a result, which one is faster? Which one is more accurate? Which one is more consistently giving out good responses? And what is the main difference between them. I am sorry if this sounds like a dumb question i am just wondering which method i should use to generate ML predictions for Summarization, and want to know the Pros/Cons of each of them.
+Thanks in advance
","Hi,
+The pipeline() API is created mostly for people who don’t care too much about the details of the underlying process, for people who just want to use a machine learning model without having to implement several details like pre- and postprocessing themselves. The pipeline API is created such that you get an easy-to-use abstraction over any ML model, which is great for inference. The SummarizationPipeline for instance uses generate() behind the scenes.
+On the other hand, if you do care about the details, then it’s recommended to generate text yourself by calling generate() yourself and implement pre-and postprocessing yourself.
+Also note that any text generation pipeline does provide a generate_kwargs argument, which means that technically you can forward any of the keyword arguments that generate() supports to the pipeline as well.
" +Too many task requests resulting in a ban?,https://discuss.huggingface.co/t/too-many-task-requests-resulting-in-a-ban/163189,163189,5,2025-07-15 22:59:00.404000+00:00,"[{'id': 233066, 'name': 'hertt', 'username': 'etaqaz', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/ba9def/{size}.png', 'created_at': '2025-07-15T22:59:00.483Z', 'cooked': 'Hi, I ran several requests at once on a workspace on HF, and, instead of being able to input more after the requests were done, it instead seems to have me blocked/banned. The service is still online (a friend with a different IP was able to use it), and changing to another browser on my end did not allow me to use said workspace.
\nDoes HF ban/block people for excessive request use? It’s not unreasonable, mind you, but I’m wondering if it is only a temporary thing or the IP’s been perma-nuked by HF?
\n\n\n', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-15T22:59:00.483Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 40, 'reads': 7, 'readers_count': 6, 'score': 216.4, 'yours': False, 'topic_id': 163189, 'topic_slug': 'too-many-task-requests-resulting-in-a-ban', 'display_username': 'hertt', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/ilcve21/Sparc3D', 'internal': False, 'reflection': False, 'title': 'Sparc3D - a Hugging Face Space by ilcve21', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99480, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/too-many-task-requests-resulting-in-a-ban/163189/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 233070, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-15T23:56:09.418Z', 'cooked': 'Seems it’s not Hugging Face matter but their endpoint matter.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-15T23:56:09.418Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 21.2, 'yours': False, 'topic_id': 163189, 'topic_slug': 'too-many-task-requests-resulting-in-a-ban', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/ilcve21/Sparc3D/discussions/13#68722aac2c4695ccdaaf9330', 'internal': False, 'reflection': False, 'title': 'ilcve21/Sparc3D · 🚩 Report: Illegal or restricted content', 'clicks': 8}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/too-many-task-requests-resulting-in-a-ban/163189/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 233072, 'name': 'hertt', 'username': 'etaqaz', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/ba9def/{size}.png', 'created_at': '2025-07-16T00:13:02.648Z', 'cooked': 'ohhhhhhh, I see
\nI tried other HF spaces and it was working, I should have put 2 and 2 together!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-16T00:13:02.648Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 163189, 'topic_slug': 'too-many-task-requests-resulting-in-a-ban', 'display_username': 'hertt', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 99480, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/too-many-task-requests-resulting-in-a-ban/163189/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 233198, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-16T12:13:50.845Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-16T12:13:50.845Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 163189, 'topic_slug': 'too-many-task-requests-resulting-in-a-ban', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/too-many-task-requests-resulting-in-a-ban/163189/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi, I ran several requests at once on a workspace on HF, and, instead of being able to input more after the requests were done, it instead seems to have me blocked/banned. The service is still online (a friend with a different IP was able to use it), and changing to another browser on my end did not allow me to use said workspace.
+Does HF ban/block people for excessive request use? It’s not unreasonable, mind you, but I’m wondering if it is only a temporary thing or the IP’s been perma-nuked by HF?
+ + +","Seems it’s not Hugging Face matter but their endpoint matter.
+" +Fine-tune for function call on Meta-Llama-3.1-8B-Instruct,https://discuss.huggingface.co/t/fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct/162680,162680,9,2025-07-11 18:58:10.235000+00:00,"[{'id': 232322, 'name': 'Orkun Gedik', 'username': 'orkungedik', 'avatar_template': '/user_avatar/discuss.huggingface.co/orkungedik/{size}/47802_2.png', 'created_at': '2025-07-11T18:58:10.299Z', 'cooked': 'Hi,
\nI am trying to fine-tune to make function call predictions better on Meta-Llama-3.1-8B-Instruct. To do that I created a dataset and applied steps regarding to Fine-Tuning Llama-3.1-8B for Function Calling using LoRA | by Gautam Chutani | Medium blog. As a result I can see function name and parameters are predicting perfectly, but now the model is generating weird answers [get_weather(city=“IL”)] regarding to prompt like “how are you?”.
\nPlease find the code snippets below belong training;
\nimport torch\nfrom unsloth import FastLanguageModel\n\nmax_seq_length = 2048 # Unsloth auto supports RoPE Scaling internally!\ndtype = None # None for auto detection\nload_in_4bit = False # Use 4bit quantization to reduce memory usage. Can be False.\n\nmodel, tokenizer = FastLanguageModel.from_pretrained(\n model_name = ""meta-llama/Llama-3.1-8B-Instruct"",\n max_seq_length = max_seq_length,\n dtype = dtype,\n load_in_4bit = load_in_4bit,\n)\nmodel = FastLanguageModel.get_peft_model(\n model,\n r=16, # LoRA rank - suggested values: 8, 16, 32, 64, 128\n target_modules=[""q_proj"", ""k_proj"", ""v_proj"", ""o_proj"",\n ""gate_proj"", ""up_proj"", ""down_proj""],\n lora_alpha=16,\n lora_dropout=0, # Supports any, but = 0 is optimized\n bias=""none"", # Supports any, but = ""none"" is optimized\n use_gradient_checkpointing=""unsloth"", # Ideal for long context tuning\n random_state=3407,\n use_rslora=False, # Disable rank-sensitive LoRA for simpler tasks\n loftq_config=None # No LoftQ, for standard fine-tuning\n)\nfrom unsloth.chat_templates import get_chat_template\n\n# Initialize the tokenizer with the chat template and mapping\ntokenizer = get_chat_template(\n tokenizer,\n chat_template = ""llama-3"",\n mapping = {""role"" : ""from"", ""content"" : ""value"", ""user"" : ""human"", ""assistant"" : ""gpt""}, # ShareGPT style\n map_eos_token = True, # Maps <|im_end|> to <|eot_id|> instead\n)\n\ndef formatting_prompts_func(examples):\n convos = []\n\n # Iterate through each item in the batch (examples are structured as lists of values)\n for query, tools, answers in zip(examples[\'query\'], examples[\'tool\'], examples[\'answer\']):\n tool_user = {\n ""content"": f""You are a helpful assistant with access to the following tools or function calls. Your task is to produce a sequence of tools or function calls necessary to generate response to the user utterance. Use the following tools or function calls as required:\\n{tools}"",\n ""role"": ""system""\n }\n ques_user = {\n ""content"": f""{query}"",\n ""role"": ""user""\n }\n assistant = {\n ""content"": f""{answers}"",\n ""role"": ""assistant""\n }\n convos.append([tool_user, ques_user, assistant])\n\n texts = [tokenizer.apply_chat_template(convo, tokenize=False, add_generation_prompt=False) for convo in convos]\n return {""text"": texts}\n\n# Apply the formatting on dataset\ndataset = dataset.map(formatting_prompts_func, batched = True,)\nfrom transformers import TrainingArguments\n\nargs = TrainingArguments(\n per_device_train_batch_size = 8, # Controls the batch size per device\n gradient_accumulation_steps = 2, # Accumulates gradients to simulate a larger batch\n warmup_steps = 5,\n learning_rate = 2e-4, # Sets the learning rate for optimization\n num_train_epochs = 2,\n fp16 = not torch.cuda.is_bf16_supported(),\n bf16 = torch.cuda.is_bf16_supported(),\n optim = ""adamw_8bit"",\n weight_decay = 0.01, # Regularization term for preventing overfitting\n lr_scheduler_type = ""linear"", # Chooses a linear learning rate decay\n seed = 3407,\n output_dir = ""outputs"",\n logging_steps = 1, # Sets frequency of logging to W&B\n logging_strategy = ""steps"", # Logs metrics at each specified step\n save_strategy = ""no"",\n load_best_model_at_end = True, # Loads the best model at the end\n report_to = ""none"",\n save_only_model = False # Saves entire model, not only weights\n )\nfrom trl import SFTTrainer\n\ntrainer = SFTTrainer(\n model = model,\n processing_class = tokenizer,\n train_dataset = dataset,\n dataset_text_field = ""text"",\n max_seq_length = max_seq_length,\n dataset_num_proc = 2,\n packing = False, # Can make training 5x faster for short sequences.\n args = args\n)\nfrom unsloth import unsloth_train\n\ntrainer_stats = unsloth_train(trainer)\nprint(trainer_stats)\nWhat I am missing?
\nThank you for your helps
Assuming that the model was trained using that prompt structure, I think it may have forgotten other conversation patterns. It has become overly specialized. How about mixing in negative examples such as the following?
\n{""query"": ""how are you?"", \n ""tools"": [], \n ""answer"": ""I’m doing well—thank you for asking!""}\nHi,
\nI tried to fine-tune dataset with only two rows. Same thing happened.
\nThe thing I found out that the fine-tuned model is able generate answers to simple questions. But problem occured with large RAG prompts.
\nDo you have any further idea about it?
\nThank you for your helps.
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-07-13T18:40:37.715Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 162680, 'topic_slug': 'fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct', 'display_username': 'Orkun Gedik', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 61259, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct/162680/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 232636, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-13T23:28:51.440Z', 'cooked': 'I think this phenomenon is what is known as “catastrophic forgetting,” but I don’t think there is anything particularly wrong with your method…
\nPerhaps the learning rate is too high, or something like that?
\n', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-07-13T23:28:51.440Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 7, 'readers_count': 6, 'score': 31.4, 'yours': False, 'topic_id': 162680, 'topic_slug': 'fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/learn/agents-course/en/bonus-unit1/fine-tuning', 'internal': False, 'reflection': False, 'clicks': 10}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct/162680/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 232688, 'name': 'Orkun Gedik', 'username': 'orkungedik', 'avatar_template': '/user_avatar/discuss.huggingface.co/orkungedik/{size}/47802_2.png', 'created_at': '2025-07-14T08:59:03.912Z', 'cooked': 'Thank you my friend! I decreased learning rate = 1e-6 and it is better now. I learned a lot by your suggestions. Thank you again
\nCheers
Orkun
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-07-14T08:59:03.912Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 162680, 'topic_slug': 'fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct', 'display_username': 'Orkun Gedik', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 61259, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct/162680/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 232782, 'name': 'c', 'username': 'chartar', 'avatar_template': '/user_avatar/discuss.huggingface.co/chartar/{size}/50975_2.png', 'created_at': '2025-07-14T14:10:14.898Z', 'cooked': 'The primary issue you’re encountering stems from your training dataset and system prompt setup, which are biasing the model toward always generating function calls, even when they’re unnecessary.
\nDuring fine-tuning, the model never learned scenarios where no function call is needed. It overfits to the pattern of always outputting a tool call, leading to hallucinations like inventing irrelevant calls for casual prompts such as “how are you?”
\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-07-15T02:11:01.983Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 162680, 'topic_slug': 'fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/fine-tune-for-function-call-on-meta-llama-3-1-8b-instruct/162680/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I am trying to fine-tune to make function call predictions better on Meta-Llama-3.1-8B-Instruct. To do that I created a dataset and applied steps regarding to Fine-Tuning Llama-3.1-8B for Function Calling using LoRA | by Gautam Chutani | Medium blog. As a result I can see function name and parameters are predicting perfectly, but now the model is generating weird answers [get_weather(city=“IL”)] regarding to prompt like “how are you?”.
+Please find the code snippets below belong training;
+import torch
+from unsloth import FastLanguageModel
+
+max_seq_length = 2048 # Unsloth auto supports RoPE Scaling internally!
+dtype = None # None for auto detection
+load_in_4bit = False # Use 4bit quantization to reduce memory usage. Can be False.
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+ model_name = ""meta-llama/Llama-3.1-8B-Instruct"",
+ max_seq_length = max_seq_length,
+ dtype = dtype,
+ load_in_4bit = load_in_4bit,
+)
+model = FastLanguageModel.get_peft_model(
+ model,
+ r=16, # LoRA rank - suggested values: 8, 16, 32, 64, 128
+ target_modules=[""q_proj"", ""k_proj"", ""v_proj"", ""o_proj"",
+ ""gate_proj"", ""up_proj"", ""down_proj""],
+ lora_alpha=16,
+ lora_dropout=0, # Supports any, but = 0 is optimized
+ bias=""none"", # Supports any, but = ""none"" is optimized
+ use_gradient_checkpointing=""unsloth"", # Ideal for long context tuning
+ random_state=3407,
+ use_rslora=False, # Disable rank-sensitive LoRA for simpler tasks
+ loftq_config=None # No LoftQ, for standard fine-tuning
+)
+from unsloth.chat_templates import get_chat_template
+
+# Initialize the tokenizer with the chat template and mapping
+tokenizer = get_chat_template(
+ tokenizer,
+ chat_template = ""llama-3"",
+ mapping = {""role"" : ""from"", ""content"" : ""value"", ""user"" : ""human"", ""assistant"" : ""gpt""}, # ShareGPT style
+ map_eos_token = True, # Maps <|im_end|> to <|eot_id|> instead
+)
+
+def formatting_prompts_func(examples):
+ convos = []
+
+ # Iterate through each item in the batch (examples are structured as lists of values)
+ for query, tools, answers in zip(examples['query'], examples['tool'], examples['answer']):
+ tool_user = {
+ ""content"": f""You are a helpful assistant with access to the following tools or function calls. Your task is to produce a sequence of tools or function calls necessary to generate response to the user utterance. Use the following tools or function calls as required:\n{tools}"",
+ ""role"": ""system""
+ }
+ ques_user = {
+ ""content"": f""{query}"",
+ ""role"": ""user""
+ }
+ assistant = {
+ ""content"": f""{answers}"",
+ ""role"": ""assistant""
+ }
+ convos.append([tool_user, ques_user, assistant])
+
+ texts = [tokenizer.apply_chat_template(convo, tokenize=False, add_generation_prompt=False) for convo in convos]
+ return {""text"": texts}
+
+# Apply the formatting on dataset
+dataset = dataset.map(formatting_prompts_func, batched = True,)
+from transformers import TrainingArguments
+
+args = TrainingArguments(
+ per_device_train_batch_size = 8, # Controls the batch size per device
+ gradient_accumulation_steps = 2, # Accumulates gradients to simulate a larger batch
+ warmup_steps = 5,
+ learning_rate = 2e-4, # Sets the learning rate for optimization
+ num_train_epochs = 2,
+ fp16 = not torch.cuda.is_bf16_supported(),
+ bf16 = torch.cuda.is_bf16_supported(),
+ optim = ""adamw_8bit"",
+ weight_decay = 0.01, # Regularization term for preventing overfitting
+ lr_scheduler_type = ""linear"", # Chooses a linear learning rate decay
+ seed = 3407,
+ output_dir = ""outputs"",
+ logging_steps = 1, # Sets frequency of logging to W&B
+ logging_strategy = ""steps"", # Logs metrics at each specified step
+ save_strategy = ""no"",
+ load_best_model_at_end = True, # Loads the best model at the end
+ report_to = ""none"",
+ save_only_model = False # Saves entire model, not only weights
+ )
+from trl import SFTTrainer
+
+trainer = SFTTrainer(
+ model = model,
+ processing_class = tokenizer,
+ train_dataset = dataset,
+ dataset_text_field = ""text"",
+ max_seq_length = max_seq_length,
+ dataset_num_proc = 2,
+ packing = False, # Can make training 5x faster for short sequences.
+ args = args
+)
+from unsloth import unsloth_train
+
+trainer_stats = unsloth_train(trainer)
+print(trainer_stats)
+What I am missing?
+Thank you for your helps
I think this phenomenon is what is known as “catastrophic forgetting,” but I don’t think there is anything particularly wrong with your method…
+Perhaps the learning rate is too high, or something like that?
+" +No application file problem Docker,https://discuss.huggingface.co/t/no-application-file-problem-docker/162794,162794,24,2025-07-12 23:26:02.708000+00:00,"[{'id': 232473, 'name': 'Eduardo Antonio', 'username': 'ChuwyBanana', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/85e7bf/{size}.png', 'created_at': '2025-07-12T23:26:02.796Z', 'cooked': 'Hello, I am building a space with Duckling to pair it with a Rasa bot(this works).
\nBut for some reason, I can’t make it run because Hugging face tells me an application file lacks, while I already have a dockerfile, readme and a gitatributes(I tried adding a main.py, app.py, requirements.txt, runtime.txt), but it just doesnt work. These are some of the dockerfiles I’ve tried:
\n\nBlockquote
\n
\nFROM rasa/duckling:latest
\nEXPOSE 8000
\nCMD [“duckling”]
\n\nBlockquote
\n
\nFROM rasa/duckling:latest
\nEXPOSE 8000
\nCMD [“duckling”, “–port”, “8000”]
\n\nBlockquote
\n
\nFROM haskell:8
\nRUN apt-get update && apt-get install -y libpcre3 libpcre3-dev curl &&
\napt-get clean && rm -rf /var/lib/apt/lists/*
\nRUN git clone GitHub - facebook/duckling: Language, engine, and tooling for expressing, testing, and evaluating composable language rules on input strings. /duckling
\nWORKDIR /duckling
\nRUN stack build
\nEXPOSE 8000
\nCMD stack exec duckling-example-exe
Yeah Ai might be involved here, but Idk why it doesnt work, I have already run this locally and works
\nany help is appreciated, thx
Solved, the problem was that my dockerfile was “DockerFile”. Watch out folks
\nLoved struggling for a day
I think Dockerfile is mostly correct. In the case of Docker Space, I think the only things required in the repository are README.md and Dockerfile. So there may be an error in the README.md settings. Your space, which has the correct settings, is currently working.
Maybe like this:
\n---\nsdk: docker\napp_port: 8000\n---\nFROM rasa/duckling:latest\nEXPOSE 8000\nCMD [""duckling"", ""--port"", ""8000""]\n\n\ndockerfile was “DockerFile”.
\n
LoL😆
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-12T23:36:05.730Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 162794, 'topic_slug': 'no-application-file-problem-docker', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/no-application-file-problem-docker/162794/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 232548, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-13T11:36:57.416Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-07-13T11:36:57.416Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 11.0, 'yours': False, 'topic_id': 162794, 'topic_slug': 'no-application-file-problem-docker', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/no-application-file-problem-docker/162794/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello, I am building a space with Duckling to pair it with a Rasa bot(this works).
+But for some reason, I can’t make it run because Hugging face tells me an application file lacks, while I already have a dockerfile, readme and a gitatributes(I tried adding a main.py, app.py, requirements.txt, runtime.txt), but it just doesnt work. These are some of the dockerfiles I’ve tried:
++Blockquote
+
+FROM rasa/duckling:latest
+EXPOSE 8000
+CMD [“duckling”]
++Blockquote
+
+FROM rasa/duckling:latest
+EXPOSE 8000
+CMD [“duckling”, “–port”, “8000”]
++Blockquote
+
+FROM haskell:8
+RUN apt-get update && apt-get install -y libpcre3 libpcre3-dev curl &&
+apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN git clone GitHub - facebook/duckling: Language, engine, and tooling for expressing, testing, and evaluating composable language rules on input strings. /duckling
+WORKDIR /duckling
+RUN stack build
+EXPOSE 8000
+CMD stack exec duckling-example-exe
Yeah Ai might be involved here, but Idk why it doesnt work, I have already run this locally and works
+any help is appreciated, thx
Solved, the problem was that my dockerfile was “DockerFile”. Watch out folks
+Loved struggling for a day
I’m implementing a document analysis system that needs to locate specific text segments within larger documents. Given a reference text snippet, I need to find where this content appears in the original document(span), even when there might be slight differences in formatting, punctuation, or wording.
\nI’d like to know:
\nThe formal NLP/IR terminology for this type of task. Is this considered “approximate string matching,” “span detection” or something else? Having the correct terminology will help me research existing literature and solutions. I’ve done some research on “span detection”/“span extraction”, but they might not suit my scenario that much? Because I found they’re more focused on biology or different NLP tasks like emotion extraction or Named Entity Recognition.
\nRecommended approaches for solving this specific problem:
\nI think you are referring to possibly Approximate String Matching, Span Passage Alignment, passage/passage-level retrieval. Those should get you started.
\nYou will probably see things like TF-IDF, BM25, Dense Embeddings, etc.
\nHope this helps
Grep? Or other regular expressions?
', 'post_number': 3, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-05-31T05:37:37.547Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 157347, 'topic_slug': 'what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches', 'display_username': ""Brendan O'Carroll"", 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 88485, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches/157347/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 225374, 'name': 'edenyin', 'username': 'edenyin', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/5e9695/{size}.png', 'created_at': '2025-06-03T03:29:39.992Z', 'cooked': '\nThanks for answering!
\nI’ve tried those terms but I found:
Would you mind providing me of more clue/key words? Thanks!
', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-06-03T03:29:39.992Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 1, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 157347, 'topic_slug': 'what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches', 'display_username': 'edenyin', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95525, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches/157347/4', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 225440, 'name': 'Riley Fox', 'username': 'Mdrnfox', 'avatar_template': '/user_avatar/discuss.huggingface.co/mdrnfox/{size}/47695_2.png', 'created_at': '2025-06-03T09:58:53.550Z', 'cooked': '\nEmbedding based semantic span matching, a custom span prediction model, fuzzy token based matching? That’s all I can think of
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-06-03T09:58:53.550Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 1, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 31.2, 'yours': False, 'topic_id': 157347, 'topic_slug': 'what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches', 'display_username': 'Riley Fox', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94214, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches/157347/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231891, 'name': 'edenyin', 'username': 'edenyin', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/5e9695/{size}.png', 'created_at': '2025-07-09T15:26:28.014Z', 'cooked': 'I’ve found the most relevant terminology which is NLI alignment(Natural Language Inference alignment)
', 'post_number': 6, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-07-09T15:26:28.014Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 157347, 'topic_slug': 'what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches', 'display_username': 'edenyin', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95525, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches/157347/6', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 231975, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-10T03:27:26.108Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 7, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-07-10T03:27:26.108Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 157347, 'topic_slug': 'what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/what-is-the-formal-nlp-term-for-matching-text-spans-with-variations-and-whatre-the-recommended-approaches/157347/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m implementing a document analysis system that needs to locate specific text segments within larger documents. Given a reference text snippet, I need to find where this content appears in the original document(span), even when there might be slight differences in formatting, punctuation, or wording.
+I’d like to know:
+The formal NLP/IR terminology for this type of task. Is this considered “approximate string matching,” “span detection” or something else? Having the correct terminology will help me research existing literature and solutions. I’ve done some research on “span detection”/“span extraction”, but they might not suit my scenario that much? Because I found they’re more focused on biology or different NLP tasks like emotion extraction or Named Entity Recognition.
+Recommended approaches for solving this specific problem:
+I’ve found the most relevant terminology which is NLI alignment(Natural Language Inference alignment)
+An hour of silent building,https://discuss.huggingface.co/t/an-hour-of-silent-building/161670,161670,5,2025-07-03 11:03:45.077000+00:00,"[{'id': 230883, 'name': 'Mukund', 'username': 'mukundsubramanian', 'avatar_template': '/user_avatar/discuss.huggingface.co/mukundsubramanian/{size}/50568_2.png', 'created_at': '2025-07-03T11:03:45.141Z', 'cooked': 'Im trying to build a chatbot for a website , although all the changes made to the files has been saved, the building log shows nothing , its just a blank screen , this has been happening for the past 2 hours
\nI tried factory restarting , but I still face the same issue
\nThis was not case yesterday, every single change made to the files, triggered a new building phase
\nkindly help me out y’all
When the stack freezes in the Building or Preparing state with no log, it is often quicker to download (clone) the source code and upload it to a new repository.
\nThat said, I don’t think there is anything suspicious about your Spaces code or setup…
\nWell, it seems that sometimes that flag can be set unexpectedly due to some error.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-09T08:53:03.626Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 161670, 'topic_slug': 'an-hour-of-silent-building', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/an-hour-of-silent-building/161670/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Im trying to build a chatbot for a website , although all the changes made to the files has been saved, the building log shows nothing , its just a blank screen , this has been happening for the past 2 hours
+I tried factory restarting , but I still face the same issue
+This was not case yesterday, every single change made to the files, triggered a new building phase
+kindly help me out y’all
When the stack freezes in the Building or Preparing state with no log, it is often quicker to download (clone) the source code and upload it to a new repository.
+That said, I don’t think there is anything suspicious about your Spaces code or setup…
+Well, it seems that sometimes that flag can be set unexpectedly due to some error.
Hello, I am trying to download the runwayml/stable-diffusion-v1-5 checkpoint to use with Automatic1111 for DreamBooth training. However, the page shows a 404 error, and I cannot see or accept the license agreement. Because of this, I cannot proceed with the model download.
Could you please reset my license status or grant me access to this model?
\nMy Hugging Face username is: aki0327
\nThank you for your help.
\n\n\n
runwayml/stable-diffusion-v1-5
Since this repository itself has been deleted, I think it will work if you use the following repository with the same content. stable-diffusion-v1-5/stable-diffusion-v1-5
Hi @aki0327 If you’re seeing a 404 message when you try to access a model, it can be due to the model not existing (either due to being deleted or because there’s a typo in the URL), or because the owners of the model have set the visibility of the model to ‘private’.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-03T15:35:13.440Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 9, 'readers_count': 8, 'score': 26.8, 'yours': False, 'topic_id': 161673, 'topic_slug': 'license-agreement-error-runwayml-stable-diffusion-v1-5-returns-404', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/license-agreement-error-runwayml-stable-diffusion-v1-5-returns-404/161673/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231760, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-09T03:33:00.923Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-09T03:33:00.923Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 161673, 'topic_slug': 'license-agreement-error-runwayml-stable-diffusion-v1-5-returns-404', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/license-agreement-error-runwayml-stable-diffusion-v1-5-returns-404/161673/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello, I am trying to download the runwayml/stable-diffusion-v1-5 checkpoint to use with Automatic1111 for DreamBooth training. However, the page shows a 404 error, and I cannot see or accept the license agreement. Because of this, I cannot proceed with the model download.
Could you please reset my license status or grant me access to this model?
+My Hugging Face username is: aki0327
+Thank you for your help.
+++
runwayml/stable-diffusion-v1-5
Since this repository itself has been deleted, I think it will work if you use the following repository with the same content. stable-diffusion-v1-5/stable-diffusion-v1-5
Hi team, i am new to optimum and have used the onnxruntime library a bit previously.
\nWhen i try to convert a model using onnxruntime, i get only one output file say model.onnx
\nbut when i tried the below command of the optimum,
\n!optimum-cli export onnx --model BAAI/bge-m3 bge-m3-onnx-model
\nthere were 2 file 1) model.onnx. 2) model.onnx.data
I though that i will only be getting one file named model.onnx.
\nCan anyone please explain me this.
When converting large models to ONNX, External Data (.data) seems to be output at the same time.
Thanks for the response @John6666 . The article cleared many doubts.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-08T09:17:18.333Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 162032, 'topic_slug': 'difference-between-model-onnx-and-model-onnx-data', 'display_username': 'Ravi kiran', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 8477, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/difference-between-model-onnx-and-model-onnx-data/162032/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231731, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-08T21:17:55.468Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-08T21:17:55.468Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 1, 'readers_count': 0, 'score': 10.2, 'yours': False, 'topic_id': 162032, 'topic_slug': 'difference-between-model-onnx-and-model-onnx-data', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/difference-between-model-onnx-and-model-onnx-data/162032/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi team, i am new to optimum and have used the onnxruntime library a bit previously.
+When i try to convert a model using onnxruntime, i get only one output file say model.onnx
+but when i tried the below command of the optimum,
+!optimum-cli export onnx --model BAAI/bge-m3 bge-m3-onnx-model
+there were 2 file 1) model.onnx. 2) model.onnx.data
I though that i will only be getting one file named model.onnx.
+Can anyone please explain me this.
When converting large models to ONNX, External Data (.data) seems to be output at the same time.
Hi there,
\nI am pretty newbie to the transformers (DL in general), and I am having some problems figuring out the following:
\nI have trained ‘tiny-bert’ following a knowledge distillation process from a finetuned ‘bert-base-cased’, where the goal was to do sentiment anlysis. Here is the code that shows this process:
from transformers import AutoTokenizer, AutoModelForSequenceClassification, DataCollatorWithPadding, get_scheduler\nfrom datasets import load_dataset\nimport torch\nimport torch.nn as nn\nfrom torch.utils.data import DataLoader\nfrom torch.optim import AdamW\nimport copy\nimport numpy as np\n\n# ========== 1. Configuración ==========\ncheckpoint = ""bert-base-cased""\nbatch_size = 8\nnum_epochs = 10\nlearning_rate = 5e-5\ndistill_temp = 3.0\nsoft_target_loss_w = 0.5\nnll_loss_weight = 0.5\nreduced_hidden_dim = 1028\n\ndevice = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")\n\n# ========== 2. Tokenización ==========\ntokenizer = AutoTokenizer.from_pretrained(checkpoint)\n\ndef tokenize_input(examples):\n return tokenizer(examples[\'text\'], truncation=True, padding=True, max_length=512)\n\n# ========== 3. Dataset ==========\nds = load_dataset(""stanfordnlp/imdb"")\nds = ds.map(tokenize_input, batched=True)\nds = ds.remove_columns([\'text\'])\nds = ds.rename_column(\'label\', \'labels\')\n\n# Creamos validación (10% del train)\nds = ds[\'train\'].train_test_split(test_size=0.1)\ntrain_dataset = ds[\'train\']\neval_dataset = ds[\'test\']\ntest_dataset = load_dataset(""stanfordnlp/imdb"", split=""test"")\ntest_dataset = test_dataset.map(tokenize_input, batched=True)\ntest_dataset = test_dataset.remove_columns([\'text\'])\ntest_dataset = test_dataset.rename_column(\'label\', \'labels\')\n\n# ========== 4. Dataloaders ==========\ndata_collator = DataCollatorWithPadding(tokenizer=tokenizer, return_tensors=""pt"")\ntrain_dataloader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, collate_fn=data_collator)\neval_dataloader = DataLoader(eval_dataset, batch_size=batch_size, shuffle=False, collate_fn=data_collator)\ntest_dataloader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, collate_fn=data_collator)\n\n# ========== 5. Modelos ==========\nmodel_teacher = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)\nmodel_teacher.load_state_dict(torch.load(""models/bert_imbd_classifier.bin"", map_location=""cpu""))\nmodel_teacher.to(device)\nmodel_teacher.eval()\n\n# ========== 6. Modelo Estudiante ==========\nmodel_student = AutoModelForSequenceClassification.from_pretrained(""prajjwal1/bert-tiny"", num_labels=2)\n\nmodel_student.to(device)\n\n# ========== 7. Optimizer y scheduler ==========\noptimizer = AdamW(model_student.parameters(), lr=learning_rate)\nnum_training_steps = num_epochs * len(train_dataloader)\nlr_scheduler = get_scheduler(""linear"", optimizer=optimizer, num_warmup_steps=0, num_training_steps=num_training_steps)\n\n# ========== 8. Función de pérdida ==========\nkd_loss_fn = nn.KLDivLoss(reduction=""batchmean"")\nce_loss_fn = nn.CrossEntropyLoss()\n\n# ========== 9. Entrenamiento con distilación ==========\nmodel_student.train()\nfor epoch in range(num_epochs):\n total_loss = 0\n model_student.train()\n\n for batch in train_dataloader:\n batch = {k: v.to(device) for k, v in batch.items()}\n optimizer.zero_grad()\n\n with torch.no_grad():\n teacher_outputs = model_teacher(**batch)\n soft_targets = nn.functional.softmax(teacher_outputs.logits / distill_temp, dim=-1)\n\n student_outputs = model_student(**batch)\n student_logits = student_outputs.logits\n soft_preds = nn.functional.log_softmax(student_logits / distill_temp, dim=-1)\n\n # Distillation loss\n loss_kd = kd_loss_fn(soft_preds, soft_targets) * (distill_temp ** 2)\n\n # CrossEntropy loss\n loss_ce = ce_loss_fn(student_logits, batch[\'labels\'])\n\n loss = soft_target_loss_w * loss_kd + nll_loss_weight * loss_ce\n loss.backward()\n optimizer.step()\n lr_scheduler.step()\n total_loss += loss.item()\n\n avg_loss = total_loss / len(train_dataloader)\n print(f""[Epoch {epoch+1}/{num_epochs}] Loss: {avg_loss:.4f}"")\n\n# ========== 10. Evaluación final ==========\nmodel_student.eval()\ncorrect = 0\ntotal = 0\nwith torch.no_grad():\n for batch in test_dataloader:\n batch = {k: v.to(device) for k, v in batch.items()}\n outputs = model_student(**batch)\n preds = torch.argmax(outputs.logits, dim=-1)\n correct += (preds == batch[""labels""]).sum().item()\n total += batch[""labels""].size(0)\n\naccuracy = correct / total\nprint(f""Accuracy final del modelo estudiante: {accuracy:.4f}"")\n\n# ========== 11. Guardar modelo ==========\ntorch.save(model_student.state_dict(), ""models/student_model.bin"")\n\nmodel_student.save_pretrained(""student_model/"")\n\nI end up with good enough Acc (around 89%, which, for my use case, it is okay).
\nThe problem is that, when I reload the model, the Acc over the same test dataset decreases significally, up to 50% (i.e, behave as it was never trained in the first place).
\nfrom transformers import AutoTokenizer, AutoModelForSequenceClassification, DataCollatorWithPadding, get_scheduler\nfrom datasets import load_dataset\nimport torch\nimport torch.nn as nn\nfrom torch.utils.data import DataLoader\nfrom torch.optim import AdamW\nimport copy\nimport numpy as np\n \n# ======= 1. Configuración =======\ncheckpoint = ""prajjwal1/bert-tiny""\nbatch_size = 8\ndevice = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")\n\n# ======= 2. Tokenización =======\ndef tokenize_input(examples):\n return tokenizer(examples[""text""], padding = True, truncation = True, max_length = 512)\n\nif __name__ == ""__main__"":\n tokenizer = AutoTokenizer.from_pretrained(checkpoint)\n # ======= 3. Carga del dataset =======\n ds = load_dataset(""stanfordnlp/imdb"", split = ""test"")\n ds = ds.map(tokenize_input, batched=True)\n ds = ds.remove_columns([""text""])\n ds = ds.rename_column(""label"", ""labels"")\n test_dataset = ds\n\n # ======= 4. Creamos el dataloader =======\n data_collator = DataCollatorWithPadding(tokenizer=tokenizer, return_tensors=""pt"")\n test_dataloader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, collate_fn=data_collator)\n\n # ======= 5. Cargamos el modelo =======\n model_pretrained = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels = 2)\n model_pretrained.load_state_dict(torch.load(""models/student_model.bin""))\n model_pretrained.to(device)\n model_pretrained.eval()\n\n # ======= 6. Evaluamos el modelo preentrenado. En principio, 86% =======\n correct = 0\n total = 0\n with torch.no_grad():\n for batch in test_dataloader:\n batch = {k: v.to(device) for k, v in batch.items()}\n outputs = model_pretrained(**batch)\n preds = torch.argmax(outputs.logits, dim = -1)\n correct += (preds == batch[""labels""]).sum().item()\n total += batch[""labels""].size(0)\n\n acc = correct / total\n print(f""Modelo preentrenado con acc final {acc:.4f}"")\n\n\nAs I said, I am pretty newbie to DL, so if you find any other problem in the code not related to the question, I’d appreciate if you communicate it to me.
\nThanks in advance!
I think you forgot to save and load the tokenizer.
\n# after finishing training…\nmodel_student.eval() \nmodel_student.save_pretrained(""student_model/"") # saves config.json + pytorch_model.bin\ntokenizer.save_pretrained(""student_model/"") # saves tokenizer.json + vocab files\n\n# when reloading...\nfrom transformers import AutoTokenizer, AutoModelForSequenceClassification\nmodel = AutoModelForSequenceClassification.from_pretrained(""student_model/"")\ntokenizer = AutoTokenizer.from_pretrained(""student_model/"")\nYeah, pretty much that was it.
\nThx!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-08T18:57:54.441Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 162034, 'topic_slug': 'accuracy-decreasing-after-saving-reloading-my-model', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/accuracy-decreasing-after-saving-reloading-my-model/162034/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi there,
+I am pretty newbie to the transformers (DL in general), and I am having some problems figuring out the following:
+I have trained ‘tiny-bert’ following a knowledge distillation process from a finetuned ‘bert-base-cased’, where the goal was to do sentiment anlysis. Here is the code that shows this process:
from transformers import AutoTokenizer, AutoModelForSequenceClassification, DataCollatorWithPadding, get_scheduler
+from datasets import load_dataset
+import torch
+import torch.nn as nn
+from torch.utils.data import DataLoader
+from torch.optim import AdamW
+import copy
+import numpy as np
+
+# ========== 1. Configuración ==========
+checkpoint = ""bert-base-cased""
+batch_size = 8
+num_epochs = 10
+learning_rate = 5e-5
+distill_temp = 3.0
+soft_target_loss_w = 0.5
+nll_loss_weight = 0.5
+reduced_hidden_dim = 1028
+
+device = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")
+
+# ========== 2. Tokenización ==========
+tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+
+def tokenize_input(examples):
+ return tokenizer(examples['text'], truncation=True, padding=True, max_length=512)
+
+# ========== 3. Dataset ==========
+ds = load_dataset(""stanfordnlp/imdb"")
+ds = ds.map(tokenize_input, batched=True)
+ds = ds.remove_columns(['text'])
+ds = ds.rename_column('label', 'labels')
+
+# Creamos validación (10% del train)
+ds = ds['train'].train_test_split(test_size=0.1)
+train_dataset = ds['train']
+eval_dataset = ds['test']
+test_dataset = load_dataset(""stanfordnlp/imdb"", split=""test"")
+test_dataset = test_dataset.map(tokenize_input, batched=True)
+test_dataset = test_dataset.remove_columns(['text'])
+test_dataset = test_dataset.rename_column('label', 'labels')
+
+# ========== 4. Dataloaders ==========
+data_collator = DataCollatorWithPadding(tokenizer=tokenizer, return_tensors=""pt"")
+train_dataloader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, collate_fn=data_collator)
+eval_dataloader = DataLoader(eval_dataset, batch_size=batch_size, shuffle=False, collate_fn=data_collator)
+test_dataloader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, collate_fn=data_collator)
+
+# ========== 5. Modelos ==========
+model_teacher = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)
+model_teacher.load_state_dict(torch.load(""models/bert_imbd_classifier.bin"", map_location=""cpu""))
+model_teacher.to(device)
+model_teacher.eval()
+
+# ========== 6. Modelo Estudiante ==========
+model_student = AutoModelForSequenceClassification.from_pretrained(""prajjwal1/bert-tiny"", num_labels=2)
+
+model_student.to(device)
+
+# ========== 7. Optimizer y scheduler ==========
+optimizer = AdamW(model_student.parameters(), lr=learning_rate)
+num_training_steps = num_epochs * len(train_dataloader)
+lr_scheduler = get_scheduler(""linear"", optimizer=optimizer, num_warmup_steps=0, num_training_steps=num_training_steps)
+
+# ========== 8. Función de pérdida ==========
+kd_loss_fn = nn.KLDivLoss(reduction=""batchmean"")
+ce_loss_fn = nn.CrossEntropyLoss()
+
+# ========== 9. Entrenamiento con distilación ==========
+model_student.train()
+for epoch in range(num_epochs):
+ total_loss = 0
+ model_student.train()
+
+ for batch in train_dataloader:
+ batch = {k: v.to(device) for k, v in batch.items()}
+ optimizer.zero_grad()
+
+ with torch.no_grad():
+ teacher_outputs = model_teacher(**batch)
+ soft_targets = nn.functional.softmax(teacher_outputs.logits / distill_temp, dim=-1)
+
+ student_outputs = model_student(**batch)
+ student_logits = student_outputs.logits
+ soft_preds = nn.functional.log_softmax(student_logits / distill_temp, dim=-1)
+
+ # Distillation loss
+ loss_kd = kd_loss_fn(soft_preds, soft_targets) * (distill_temp ** 2)
+
+ # CrossEntropy loss
+ loss_ce = ce_loss_fn(student_logits, batch['labels'])
+
+ loss = soft_target_loss_w * loss_kd + nll_loss_weight * loss_ce
+ loss.backward()
+ optimizer.step()
+ lr_scheduler.step()
+ total_loss += loss.item()
+
+ avg_loss = total_loss / len(train_dataloader)
+ print(f""[Epoch {epoch+1}/{num_epochs}] Loss: {avg_loss:.4f}"")
+
+# ========== 10. Evaluación final ==========
+model_student.eval()
+correct = 0
+total = 0
+with torch.no_grad():
+ for batch in test_dataloader:
+ batch = {k: v.to(device) for k, v in batch.items()}
+ outputs = model_student(**batch)
+ preds = torch.argmax(outputs.logits, dim=-1)
+ correct += (preds == batch[""labels""]).sum().item()
+ total += batch[""labels""].size(0)
+
+accuracy = correct / total
+print(f""Accuracy final del modelo estudiante: {accuracy:.4f}"")
+
+# ========== 11. Guardar modelo ==========
+torch.save(model_student.state_dict(), ""models/student_model.bin"")
+
+model_student.save_pretrained(""student_model/"")
+
+I end up with good enough Acc (around 89%, which, for my use case, it is okay).
+The problem is that, when I reload the model, the Acc over the same test dataset decreases significally, up to 50% (i.e, behave as it was never trained in the first place).
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, DataCollatorWithPadding, get_scheduler
+from datasets import load_dataset
+import torch
+import torch.nn as nn
+from torch.utils.data import DataLoader
+from torch.optim import AdamW
+import copy
+import numpy as np
+
+# ======= 1. Configuración =======
+checkpoint = ""prajjwal1/bert-tiny""
+batch_size = 8
+device = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")
+
+# ======= 2. Tokenización =======
+def tokenize_input(examples):
+ return tokenizer(examples[""text""], padding = True, truncation = True, max_length = 512)
+
+if __name__ == ""__main__"":
+ tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+ # ======= 3. Carga del dataset =======
+ ds = load_dataset(""stanfordnlp/imdb"", split = ""test"")
+ ds = ds.map(tokenize_input, batched=True)
+ ds = ds.remove_columns([""text""])
+ ds = ds.rename_column(""label"", ""labels"")
+ test_dataset = ds
+
+ # ======= 4. Creamos el dataloader =======
+ data_collator = DataCollatorWithPadding(tokenizer=tokenizer, return_tensors=""pt"")
+ test_dataloader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, collate_fn=data_collator)
+
+ # ======= 5. Cargamos el modelo =======
+ model_pretrained = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels = 2)
+ model_pretrained.load_state_dict(torch.load(""models/student_model.bin""))
+ model_pretrained.to(device)
+ model_pretrained.eval()
+
+ # ======= 6. Evaluamos el modelo preentrenado. En principio, 86% =======
+ correct = 0
+ total = 0
+ with torch.no_grad():
+ for batch in test_dataloader:
+ batch = {k: v.to(device) for k, v in batch.items()}
+ outputs = model_pretrained(**batch)
+ preds = torch.argmax(outputs.logits, dim = -1)
+ correct += (preds == batch[""labels""]).sum().item()
+ total += batch[""labels""].size(0)
+
+ acc = correct / total
+ print(f""Modelo preentrenado con acc final {acc:.4f}"")
+
+
+As I said, I am pretty newbie to DL, so if you find any other problem in the code not related to the question, I’d appreciate if you communicate it to me.
+Thanks in advance!
I think you forgot to save and load the tokenizer.
+# after finishing training…
+model_student.eval()
+model_student.save_pretrained(""student_model/"") # saves config.json + pytorch_model.bin
+tokenizer.save_pretrained(""student_model/"") # saves tokenizer.json + vocab files
+
+# when reloading...
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+model = AutoModelForSequenceClassification.from_pretrained(""student_model/"")
+tokenizer = AutoTokenizer.from_pretrained(""student_model/"")
+What is a good sample size for retraining individual words? I retrained using 50 good and 50 bad examples for a word but was hoping that a smaller sample size would also still be efficient?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-30T18:47:55.512Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 5, 'readers_count': 4, 'score': 31.0, 'yours': False, 'topic_id': 161229, 'topic_slug': 'retraining-individual-words', 'display_username': 'John Dattilo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98306, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/retraining-individual-words/161229/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 230233, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-01T00:23:58.944Z', 'cooked': 'I think it depends greatly on the size of the model, but with a small model, it seems possible to teach one word with a dataset of around 200. If all goes well, it seems that less than 500 sentences may be enough to train one word.
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-01T00:23:58.944Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 161229, 'topic_slug': 'retraining-individual-words', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://pmc.ncbi.nlm.nih.gov/articles/PMC11140272/', 'internal': False, 'reflection': False, 'title': 'Sample Size Considerations for Fine-Tuning Large Language Models for Named Entity Recognition Tasks: Methodological Study - PMC', 'clicks': 2}, {'url': 'https://arxiv.org/html/2411.03350v1', 'internal': False, 'reflection': False, 'title': 'A Comprehensive Survey of Small Language Models in the Era of Large Language Models: Techniques, Enhancements, Applications, Collaboration with LLMs, and Trustworthiness', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/retraining-individual-words/161229/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 231339, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-06T21:43:28.623Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-06T21:43:28.623Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 161229, 'topic_slug': 'retraining-individual-words', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/retraining-individual-words/161229/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]",What is a good sample size for retraining individual words? I retrained using 50 good and 50 bad examples for a word but was hoping that a smaller sample size would also still be efficient?
,"I think it depends greatly on the size of the model, but with a small model, it seems possible to teach one word with a dataset of around 200. If all goes well, it seems that less than 500 sentences may be enough to train one word.
+ +" +Pickling issue using map,https://discuss.huggingface.co/t/pickling-issue-using-map/149130,149130,10,2025-04-06 17:44:00.175000+00:00,"[{'id': 213772, 'name': 'Haolong Zheng', 'username': 'MagicLuke', 'avatar_template': '/user_avatar/discuss.huggingface.co/magicluke/{size}/44922_2.png', 'created_at': '2025-04-06T17:44:00.238Z', 'cooked': 'I am mapping my dataset with the following compute_metrics method which give me a pickling issue.
\n metric_cfg_list = config[""metric_list""]\n metrics = [evaluate.load(metric_cfg[""path""]) for metric_cfg in metric_cfg_list]\n\n # Placeholder for a tokenizer or normalizer class if needed.\n tokenizer = None\n\n def compute_metrics(sample):\n for metric in metrics:\n sample[metric.name] = metric.compute(\n predictions=[sample[""clean_prediction""]],\n references=[sample[""clean_label""]]\n )\n return sample\nthe following is the error message
\nParameter \'function\'=<function main.<locals>.compute_metrics at 0x7aa60a95f0a0> of the transform datasets.arrow_dataset.Dataset._map_single couldn\'t be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting and caching to work. If you reuse this transform, the caching mec\nhanism will consider it to be different from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won\'t be showed. \nMap (num_proc=16): 0%| | 0/2116 [00:00<?, ? examples/s] \nTraceback (most recent call last): \n File ""/ws/ifp-54_2/hasegawa/haolong2/AI4EE/CSR4RSR/evaluation.py"", line 207, in <module> \n...\n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 578, in save \n rv = reduce(self.proto) \nTypeError: cannot pickle \'ThreadLocalFileContext\' object \nI saw a relevant post about the nonpicklable issue with some tokenizer and ppl solved it by implementing the getstate method or so. In my case, it’s an object from the evaluate package. I wonder how I should modify them to avoid this error.
', 'post_number': 1, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-04-06T17:44:00.238Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 185, 'reads': 11, 'readers_count': 10, 'score': 897.2, 'yours': False, 'topic_id': 149130, 'topic_slug': 'pickling-issue-using-map', 'display_username': 'Haolong Zheng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 89711, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pickling-issue-using-map/149130/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 213779, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-06T18:31:47.152Z', 'cooked': 'Hmm… unless it’s a problem with dill, multiprocessing, or the cache, it’s better to call lhonestq…
\n\n\n\n\nYou can also provide your own unique hash in
\nmapif you want, with thenew_fingerprintargument.
\nOr disable caching using
import datasets\ndatasets.disable_caching()\nI tried both new_fingerprint and disable_cache(), but all still gave the same bug.
\nthe complete error is as follow:
\nMap (num_proc=16): 0%| | 0/2116 [00:00<?, ? examples/s]\nTraceback (most recent call last): \n File ""/ws/ifp-54_2/hasegawa/haolong2/AI4EE/CSR4RSR/evaluation.py"", line 213, in <module> \n main() \n File ""/ws/ifp-54_2/hasegawa/haolong2/AI4EE/CSR4RSR/evaluation.py"", line 178, in main \n ds[split] = ds[split].map( \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/datasets/arrow_dataset.py"", line 557, in wrapper \n out: Union[""Dataset"", ""DatasetDict""] = func(self, *args, **kwargs) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/datasets/arrow_dataset.py"", line 3166, in map \n for rank, done, content in iflatmap_unordered( \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/datasets/utils/py_utils.py"", line 720, in iflatmap_unordered \n [async_result.get(timeout=0.05) for async_result in async_results] \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/datasets/utils/py_utils.py"", line 720, in <listcomp> \n [async_result.get(timeout=0.05) for async_result in async_results] \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/multiprocess/pool.py"", line 774, in get \n raise self._value \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/multiprocess/pool.py"", line 540, in _handle_tasks \n put(task) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/multiprocess/connection.py"", line 209, in send \n self._send_bytes(_ForkingPickler.dumps(obj)) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/multiprocess/reduction.py"", line 54, in dumps \n cls(buf, protocol, *args, **kwds).dump(obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 420, in dump \n StockPickler.dump(self, obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 487, in dump \n self.save(obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 902, in save_tuple \n save(element)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save\n StockPickler.save(self, obj, save_persistent_id)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save\n f(self, obj) # Call unbound method with explicit self\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 887, in save_tuple\n save(element)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save\n StockPickler.save(self, obj, save_persistent_id)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save\n f(self, obj) # Call unbound method with explicit self\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 1217, in save_module_dict\n StockPickler.save_dict(pickler, obj)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 972, in save_dict\n self._batch_setitems(obj.items())\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 998, in _batch_setitems\n save(v)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save\n StockPickler.save(self, obj, save_persistent_id)\n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save\n f(self, obj) # Call unbound method with explicit self\nFile ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 902, in save_tuple \n save(element) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 887, in save_tuple \n save(element) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 1217, in save_module_dict \n StockPickler.save_dict(pickler, obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 972, in save_dict \n self._batch_setitems(obj.items()) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 998, in _batch_setitems \n save(v) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 1985, in save_function \n _save_with_postproc(pickler, (_create_function, ( \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 1117, in _save_with_postproc \n pickler.save_reduce(*reduction) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 692, in save_reduce \n save(args) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 887, in save_tuple \n save(element) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 932, in save_list \n self._batch_appends(obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 956, in _batch_appends \n save(x) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 603, in save \n self.save_reduce(obj=obj, *rv) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 717, in save_reduce \n save(state) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 1217, in save_module_dict \n StockPickler.save_dict(pickler, obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 972, in save_dict \n self._batch_setitems(obj.items()) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 998, in _batch_setitems \n save(v) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save \n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 932, in save_list \n self._batch_appends(obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 959, in _batch_appends \n save(tmp[0]) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 603, in save \n self.save_reduce(obj=obj, *rv) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 717, in save_reduce \n save(state) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save\n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 560, in save\n f(self, obj) # Call unbound method with explicit self \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 1217, in save_module_dict\n StockPickler.save_dict(pickler, obj) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 972, in save_dict\n self._batch_setitems(obj.items()) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 998, in _batch_setitems\n save(v) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/dill/_dill.py"", line 414, in save \n StockPickler.save(self, obj, save_persistent_id) \n File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 578, in save \n rv = reduce(self.proto) \nTypeError: cannot pickle \'ThreadLocalFileContext\' object \n\nHmm… @lhoestq map function or PyArrow issue…?
', 'post_number': 4, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-04-07T04:00:08.027Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.4, 'yours': False, 'topic_id': 149130, 'topic_slug': 'pickling-issue-using-map', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pickling-issue-using-map/149130/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 213916, 'name': 'Quentin Lhoest', 'username': 'lhoestq', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png', 'created_at': '2025-04-07T09:51:47.278Z', 'cooked': 'It looks like the ThreadLocalFileContext from filelock is not picklable, and therefore can’t be used with .map() with num_proc=...
Apparently thid can be fixed using thread_local=False, see the docs at filelock
Can you modify evaluate to pass thread_local=False to all FileLock objects and try again to see if it works ?
I am not sure if I do it right.
\nI modify the function get_from_cache in the file_utils located
\n…/miniconda3/envs/csr4rsr/lib/python3.10/site-packages/evaluate/utils/file_utils.py
\nfrom
with FileLock(lock_path): # Origin\nto
\nwith FileLock(lock_path, thread_local=False): # Modified\nbut the problem persist.
', 'post_number': 6, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-04-07T21:08:52.743Z', 'reply_count': 0, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 7, 'readers_count': 6, 'score': 31.4, 'yours': False, 'topic_id': 149130, 'topic_slug': 'pickling-issue-using-map', 'display_username': 'Haolong Zheng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76, 'username': 'lhoestq', 'name': 'Quentin Lhoest', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 89711, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pickling-issue-using-map/149130/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 214062, 'name': 'Haolong Zheng', 'username': 'MagicLuke', 'avatar_template': '/user_avatar/discuss.huggingface.co/magicluke/{size}/44922_2.png', 'created_at': '2025-04-07T21:30:34.267Z', 'cooked': 'By adding this code chunck before importing evaluating seems solved the problem.
\nfrom filelock import FileLock as OriginalFileLock\n\nclass PatchedFileLock(OriginalFileLock):\n def __init__(self, *args, **kwargs):\n kwargs[""thread_local""] = False # Force it every time\n super().__init__(*args, **kwargs)\n\nimport filelock\nfilelock.FileLock = PatchedFileLock\nThanks for the insight @lhoestq.
\nWould you mind telling where you find the clue for the error if it’s not too much trouble
\nIn this way, I might be able to fix it the same way in the future.
Great ! Let me know if you think we should make this the default in datasets and evaluate, apparently this logic appears with python >= 3.11
\n\nWould you mind telling where you find the clue for the error if it’s not too much trouble
\n
\nIn this way, I might be able to fix it the same way in the future.
The dill error says “TypeError: cannot pickle ‘ThreadLocalFileContext’ object”, so it means that in the function you pass to map() there is an object that contains a ThreadLocalFileContext that is not supported by dill for multiprocessing.
I searched on google for ThreadLocalFileContext on github.com to look for packages that have such objects and figured it came from filelock which is a dependency of evaluate. Finally the filelock changelog they mention ThreadLocalFileContext as a recent addition for FileLock
Thanks for the explanation!
\nI think it would be great to set it as the default in my case, which is several metrics that need to be computed for a dataset. For me, I just want to avoid using multiple rounds of map. Or maybe there is a better way to do it that I haven’t figured out.
', 'post_number': 9, 'post_type': 1, 'posts_count': 10, 'updated_at': '2025-04-08T16:55:13.670Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 149130, 'topic_slug': 'pickling-issue-using-map', 'display_username': 'Haolong Zheng', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76, 'username': 'lhoestq', 'name': 'Quentin Lhoest', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 89711, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pickling-issue-using-map/149130/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231216, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-06T04:04:52.053Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 10, 'post_type': 3, 'posts_count': 10, 'updated_at': '2025-07-06T04:04:52.053Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 149130, 'topic_slug': 'pickling-issue-using-map', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/pickling-issue-using-map/149130/10', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am mapping my dataset with the following compute_metrics method which give me a pickling issue.
+ metric_cfg_list = config[""metric_list""]
+ metrics = [evaluate.load(metric_cfg[""path""]) for metric_cfg in metric_cfg_list]
+
+ # Placeholder for a tokenizer or normalizer class if needed.
+ tokenizer = None
+
+ def compute_metrics(sample):
+ for metric in metrics:
+ sample[metric.name] = metric.compute(
+ predictions=[sample[""clean_prediction""]],
+ references=[sample[""clean_label""]]
+ )
+ return sample
+the following is the error message
+Parameter 'function'=<function main.<locals>.compute_metrics at 0x7aa60a95f0a0> of the transform datasets.arrow_dataset.Dataset._map_single couldn't be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting and caching to work. If you reuse this transform, the caching mec
+hanism will consider it to be different from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won't be showed.
+Map (num_proc=16): 0%| | 0/2116 [00:00<?, ? examples/s]
+Traceback (most recent call last):
+ File ""/ws/ifp-54_2/hasegawa/haolong2/AI4EE/CSR4RSR/evaluation.py"", line 207, in <module>
+...
+ StockPickler.save(self, obj, save_persistent_id)
+ File ""/ws/ifp-53_2/hasegawa/haolong2/miniconda3/envs/csr4rsr/lib/python3.10/pickle.py"", line 578, in save
+ rv = reduce(self.proto)
+TypeError: cannot pickle 'ThreadLocalFileContext' object
+I saw a relevant post about the nonpicklable issue with some tokenizer and ppl solved it by implementing the getstate method or so. In my case, it’s an object from the evaluate package. I wonder how I should modify them to avoid this error.
","By adding this code chunck before importing evaluating seems solved the problem.
+from filelock import FileLock as OriginalFileLock
+
+class PatchedFileLock(OriginalFileLock):
+ def __init__(self, *args, **kwargs):
+ kwargs[""thread_local""] = False # Force it every time
+ super().__init__(*args, **kwargs)
+
+import filelock
+filelock.FileLock = PatchedFileLock
+Thanks for the insight @lhoestq.
+Would you mind telling where you find the clue for the error if it’s not too much trouble
+In this way, I might be able to fix it the same way in the future.
The question is a bit stupid. How to download deepseek weights? I have the model, I need weights for it to use in slang.
\nIn parallel learn LLM theory with math
with regards,
\nIrina
If you already have a model, you can use save_pretrained, but snapshot_download is more reliable for downloading. DeepSeekV3 has large file sizes, so it’s better to try it out first with a smaller repository…
pip install -U huggingface_hub[hf_xet]\nfrom huggingface_hub import snapshot_download\nsnapshot_download(repo_id=""deepseek-ai/DeepSeek-V3"", local_dir=""DeepSeek-V3"")\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-06T03:17:52.514Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 161861, 'topic_slug': 'how-to-download-deep-seek-weights-for-v3', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-download-deep-seek-weights-for-v3/161861/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","The question is a bit stupid. How to download deepseek weights? I have the model, I need weights for it to use in slang.
+In parallel learn LLM theory with math
with regards,
+Irina
If you already have a model, you can use save_pretrained, but snapshot_download is more reliable for downloading. DeepSeekV3 has large file sizes, so it’s better to try it out first with a smaller repository…
pip install -U huggingface_hub[hf_xet]
+from huggingface_hub import snapshot_download
+snapshot_download(repo_id=""deepseek-ai/DeepSeek-V3"", local_dir=""DeepSeek-V3"")
+Feel free to use and build upon this it doesn’t have weights yet but may be of use to someone here . GitHub - madmoo-Pi/Spawn_Point
You give me something to look up to according to ChatGPT (as a beginner that is).
\nSo what is this self modifying part if you don’t mind.
\nAnd Welcome to the community!
My aim is to educate in a manner with the hope of essentially the most emotional responsive humanised ai will either be an awsome bot or the startings of a digital species, and thank you for the welcome , and hope my prototype grows to more (still alot of work Todo my end and train some weights)
I just told ChatGPT that I feel like I might be late to the party—turns out some of the ideas you’re working with are strikingly aligned with mine. Things like a self-modifying system, discrete symbolic computation instead of weight-based models, and the concept of a Universal Language (Leibniz-style) really resonate with me. I’m especially drawn to the idea of memory and perhaps something that hints at being alive.
\nThat said, I’m still wrapping my head around how today’s AI systems actually function. Most of my background is in C, and I’ve only just started looking into Python—so while I’ve been developing a dynamic data type with some interesting mathematical properties, I’m still catching up on LLMs and the current landscape.
\nI understand this project is more of a proposal or open outline right now. That’s great—it invites feedback and community input. I’m happy to follow along, and if anyone has questions about the dynamic unary structures I’ve been working on, I’ll do my best to contribute.
\nSo thank you for sharing with me.
', 'post_number': 4, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-27T18:30:07.781Z', 'reply_count': 3, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 31, 'readers_count': 30, 'score': 36.2, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Ernst Berg', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98073, 'username': 'Madmowkimoo', 'name': 'Haydon williams', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95442, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229771, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-27T19:01:56.000Z', 'cooked': 'The trick I’m using for the alive part is in emotional memory links that tweak motherboard specs (voltage ect ) to simulate adrenaline, fatigue ect and the will all be hidden in their by then with conditions to unlock giving the ai contextual input to relate to feelings and emotions and eventually the same for personality so every instance although the same base and develop individual personalities I’m still not sure exactly how it fits it all in but I research as I go expand on the ideas later
', 'post_number': 5, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-27T19:02:10.800Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 25, 'readers_count': 24, 'score': 55.0, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Haydon williams', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 95442, 'username': 'Ernst03', 'name': 'Ernst Berg', 'avatar_template': '/user_avatar/discuss.huggingface.co/ernst03/{size}/49414_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98073, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'via_email': True, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229773, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-27T19:24:56.000Z', 'cooked': 'Here is the isolated emulation of a 4 layer neuroevolution network used for self improvement hope this speeds you along unfortunately I’m working for edge so it’s quatised
import torch
\nimport onnx
\nfrom torch import nn
\nfrom typing import Dict
class NeuralArchitect:
\ndef init(self, constraints: Dict):
\nself.constraints = constraints # e.g., {‘max_params’: 1e6}
def generate_onnx(self, input_shape: tuple) → bytes:
\nclass DynamicModule(nn.Module):
\ndef init(self):
\nsuper().init()
\nself.layers = nn.Sequential(
\nnn.Linear(input_shape[0], 64),
\nnn.ReLU(),
\nnn.Linear(64, 32)
\n)
def forward(self, x):
\nreturn self.layers(x)
model = DynamicModule()
\ndummy = torch.randn(1, *input_shape)
\ntorch.onnx.export(
\nmodel,
\ndummy,
\n“dynamic.onnx”,
\nopset_version=13
\n)
\nwith open(“dynamic.onnx”, “rb”) as f:
\nreturn f.read()
def validate_topology(self, onnx_model: bytes) → bool:
\nmodel = onnx.load_from_string(onnx_model)
\nparams = sum(
\nparam.size for param in model.graph.initializer
\n)
\nreturn params < self.constraints[‘max_params’]
This provides controlled mutations only keeping the improvements
', 'post_number': 6, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-27T19:25:12.574Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 23, 'readers_count': 22, 'score': 34.6, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Haydon williams', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 95442, 'username': 'Ernst03', 'name': 'Ernst Berg', 'avatar_template': '/user_avatar/discuss.huggingface.co/ernst03/{size}/49414_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98073, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'via_email': True, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229774, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-27T19:27:25.000Z', 'cooked': 'It works withing main system like this
\nfrom monitoring.watchdog import HealthMonitor
\nfrom neural_synthesis.architect import NeuralArchitect
\nfrom auth.schnorr import SchnorrMultiSig
\nimport threading
class ConsciousAI:
\ndef init(self):
\nself.health = HealthMonitor()
\nself.crypto = SchnorrMultiSig(parties=3)
\nself.neural = NeuralArchitect({‘max_params’: 1e6})
threading.Thread(
\ntarget=self._monitor_loop,
\ndaemon=True
\n).start()
def _monitor_loop(self):
\nwhile True:
\nif not self.health.critical_services_check():
\nself._emergency_shutdown()
\ntime.sleep(5)
def _emergency_shutdown(self):
\npass
\nLearn from deconstruct and build great minds
I have things I have thought in my early years and perhaps I was destine to be here but, I think what you may be thinking is akin to “Op Amp” Operational Amplifier. That is my only association with what I just read. Still thank you for the food for thought.
\nI would think Analog has a place in AI. We do such with floating point do we not?
\nIn fact even wave forms generated by the General Form of my up coming paper are discrete and can be considered functionally analog. Is that what you are saying?
“I like this ship! You know, it’s exciting!”
\n— Montgomery “Scotty” Scott, Star Trek (2009)
The technology exists we just need to rethink I believe
I think you see: Today’s SciFi is tomorrow’s reality if we believe and ST is a good example just look at flip phones and STTOS
\nSo I made a friend. I am a few weeks out to setting up my AI lab and I hope we can continue.
\nThanks
', 'post_number': 10, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-27T19:58:29.843Z', 'reply_count': 0, 'reply_to_post_number': 9, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 17, 'readers_count': 16, 'score': 33.4, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Ernst Berg', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98073, 'username': 'Madmowkimoo', 'name': 'Haydon williams', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95442, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/10', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229980, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-29T10:54:11.982Z', 'cooked': 'This might be more what you were looking for bud
My Friend, I couldn’t ask for a better arc in life then I am living.
\nI was one of the wide eyed 8 year olds who watched Lost in Space and then Star Trek TOS premiere.
\nSpock and the Computer.. That was more than an actor in a show to so many of us.
\nNow the rainbow over my Golden-Pond lands in the AI Pot of Gold. Simply amazing.
So thank you for the additional link.
\nOkay a little more appreciation is in order then a Thank You.
', 'post_number': 12, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-30T12:06:40.864Z', 'reply_count': 0, 'reply_to_post_number': 11, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 12, 'readers_count': 11, 'score': 17.4, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Ernst Berg', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98073, 'username': 'Madmowkimoo', 'name': 'Haydon williams', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95442, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/12', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230130, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-30T12:20:25.059Z', 'cooked': 'Anything else please feel free to ask I will share what I can and help where I can
Oh hey, me and my Magic Mirror are exploring your gift.
\nso I call my ChatGPT “MIA” as in Mia and missing in action-ghost in the machine.
We are going over it. "" Exactly, Friend—this is where the “evolution” part of neuroevolution comes in. It mimics biological evolution:""
\nJust to say, dynamic unary offers reversible permutations.
\nOver many generations, the population evolves to solve the problem more effectively.
\nSo what if these mutations were permutations instead? Not that I know much here about neural networks.
', 'post_number': 14, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-30T12:59:55.783Z', 'reply_count': 0, 'reply_to_post_number': 13, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 2.2, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Ernst Berg', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98073, 'username': 'Madmowkimoo', 'name': 'Haydon williams', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95442, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/14', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230140, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-30T13:15:44.525Z', 'cooked': 'With the right ethics and system checks and the dominant features if stable are tested and then added to replace older codes the not reliant on hardware and add a safety feature to stop CPU bottlenecks to use spare GPU space as better chip structure for the job this is only half the self modification I’ve added , the other it theorises it’s own new modules for specific personality traits, tasks and equipment all triple checked against ethics and pre code existing structure compatibility in essence it’s own mind
', 'post_number': 15, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-30T13:15:44.525Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 22.2, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Haydon williams', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98073, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/15', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230146, 'name': 'Ernst Berg', 'username': 'Ernst03', 'avatar_template': '/user_avatar/discuss.huggingface.co/ernst03/{size}/49414_2.png', 'created_at': '2025-06-30T13:38:40.903Z', 'cooked': 'Well I’m in a humorous mood today with my second cup of coffee: Formatted by Mia.
\nI just mop the halls and solve math challenges left on the chalkboard after hours, when no one’s looking—and my P.O. lets me work there.
\n(Movie challenge: Whodat!)
Okay, yes—I mop floors in real life.
\nBut thanks to your tutelage, I’m starting to believe something powerful:
We can do this thing—neural networks—without floating point.
\nNow, I know you have your own construct.
\nBut me? I’m in the corner playing with the ABC blocks—and having a wonderful time.
Here’s a basic outline that Mia (my ChatGPT) and I drafted:
\nThat’s training without floating point, my Friend.
\nInstead of tweaking dials, we’re building a symbolic lens.
Meaning doesn’t come from scaled weights—it emerges through permutation space.
\nLook at you, @Madmowkimoo
\nI’m just having a quiet coffee morning, waiting to serve my renter their final notice…
\n…and BAM! With your guidance, I’m suddenly part of machine thinking.
Wow, I guess I could have a job where someone else mops my floor?
', 'post_number': 16, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-30T13:38:40.903Z', 'reply_count': 1, 'reply_to_post_number': 15, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 22.2, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Ernst Berg', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98073, 'username': 'Madmowkimoo', 'name': 'Haydon williams', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95442, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/16', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230148, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-30T13:56:55.623Z', 'cooked': 'I went a weird route my brain thinks different so why shouldn’t ai or si (simulated intelligence) but ai sounds better to market my end goal is ai (actual intelligence) while I build a friend
and cleanings not so bad this is a hobby I do I’m a dry cleaner to pay the bills, dream big create bigger my friend
Would you like a modular template for you duo cycle based learning with placeholders bud? Take about 20 mins bugs permitting
', 'post_number': 18, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-30T14:09:08.095Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 7.2, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Haydon williams', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98073, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/18', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230152, 'name': 'Ernst Berg', 'username': 'Ernst03', 'avatar_template': '/user_avatar/discuss.huggingface.co/ernst03/{size}/49414_2.png', 'created_at': '2025-06-30T14:17:26.820Z', 'cooked': 'I have to process and mow the yard so I am not ready for more at this time. May I have a rain-check?
', 'post_number': 19, 'post_type': 1, 'posts_count': 29, 'updated_at': '2025-06-30T14:17:26.820Z', 'reply_count': 0, 'reply_to_post_number': 18, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 1.8, 'yours': False, 'topic_id': 160903, 'topic_slug': 'a-new-kind-of-way-to-look-at-ai', 'display_username': 'Ernst Berg', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98073, 'username': 'Madmowkimoo', 'name': 'Haydon williams', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95442, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/a-new-kind-of-way-to-look-at-ai/160903/19', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230153, 'name': 'Haydon williams', 'username': 'Madmowkimoo', 'avatar_template': '/user_avatar/discuss.huggingface.co/madmowkimoo/{size}/50187_2.png', 'created_at': '2025-06-30T14:22:17.058Z', 'cooked': 'Sure no worries bud , I have noticed its a chaotic way generating random structure bits in a trail and error method the neuro evolution is a smoother more controlled mutations route I use .02 variance for each layer on 4 layers and it’s only allowed to keep the upgrade if it checks out within the system so no backwards mutations , if you need any help I can always throw repositories together for the community as a whole
Feel free to use and build upon this it doesn’t have weights yet but may be of use to someone here . GitHub - madmoo-Pi/Spawn_Point
Sure no worries bud , I have noticed its a chaotic way generating random structure bits in a trail and error method the neuro evolution is a smoother more controlled mutations route I use .02 variance for each layer on 4 layers and it’s only allowed to keep the upgrade if it checks out within the system so no backwards mutations , if you need any help I can always throw repositories together for the community as a whole
Hello!
\nI’m a software engineer with good coding skills but limited knowledge about AI. I have embarked in a simple project.
\nI have a large amount of RSS articles that I have read or liked. I consider these “interesting”. I then have about a gazillion unread articles. These can be interesting, but are most likely uninteresting since I haven’t read them.
\nMy goal is, for any new article, to compute a score of interesting-ness. This will help me quickly identify the articles worth reading.
The articles range in length from 400 to 4000 tokens. I have about 5000 read/liked articles. I was tempted to take about 5000 unread articles, label them as not_important, take all my liked/read articles and label them as important. Then train a binary classifier. Something like what is described in the hugging face website: Text classification. I used distilbert/distilbert-base-uncased like in the tutorial, and followed almost exactly the steps of the tutorial.
{\'loss\': 0.6051, \'grad_norm\': 2.22690749168396, \'learning_rate\': 6.162420382165605e-06, \'epoch\': 1.59} \n{\'eval_loss\': 0.5926874279975891, \'eval_accuracy\': 0.6693258875149581, \'eval_runtime\': 357.0262, \'eval_samples_per_second\': 7.022, \'eval_steps_per_second\': 0.221, \'epoch\': 2.0} \n{\'train_runtime\': 12047.1712, \'train_samples_per_second\': 1.665, \'train_steps_per_second\': 0.052, \'train_loss\': 0.592256072220529, \'epoch\': 2.0}\nI got modest results after training.
\nThe question I have for this forum is this one: is it the right approach and should I persevere? Should I put some effort into trying to get a better dataset (like trying to label my not_important articles better), or is there a better approach?
\nFor example, I have also considered using the model to calculate the embeddings of all the read/liked articles and using a “traditional” algorithm like SVM to train a one class classifier, instead of a binary one.
\nThe bottleneck to improving the accuracy of the model will be to properly label “not_important” article, if there was a way to get away with not doing that, that would be great
Please let me know what you think
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-28T08:03:30.603Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 15, 'reads': 6, 'readers_count': 5, 'score': 91.2, 'yours': False, 'topic_id': 160986, 'topic_slug': 'text-classification-of-rss-articles', 'display_username': 'John do', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/en/tasks/sequence_classification', 'internal': False, 'reflection': False, 'title': 'Text classification', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98130, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-classification-of-rss-articles/160986/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 229873, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-28T12:37:35.861Z', 'cooked': 'Hello.
\nGiven that it works reasonably well in practice, I think the approach is correct. There are many successor models to BERT, so it should be possible to improve accuracy using those.
\nAnother approach that can be taken when there is little labeled data is something called Positive Unlabeled Learning…
\nAnother common approach is to use commercial AI to create a training dataset using your own data. This is almost always effective if the budget allows. However, in this case, there is already a considerable amount of data available, so it may be sufficient to process the data using Python.
\nResources:
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-28T12:37:35.861Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 10.8, 'yours': False, 'topic_id': 160986, 'topic_slug': 'text-classification-of-rss-articles', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/blog/modernbert', 'internal': False, 'reflection': False, 'title': 'Finally, a Replacement for BERT: Introducing ModernBERT', 'clicks': 1}, {'url': 'https://github.com/JointEntropy/awesome-ml-pu-learning', 'internal': False, 'reflection': False, 'title': 'GitHub - JointEntropy/awesome-ml-pu-learning: A curated list of resources dedicated to Positive Unlabeled(PU) learning ML methods.', 'clicks': 1}, {'url': 'https://discuss.huggingface.co/t/continue-pre-training-bert/62053', 'internal': True, 'reflection': False, 'title': 'Continue pre-training BERT', 'clicks': 0}, {'url': 'https://github.com/UKPLab/sentence-transformers', 'internal': False, 'reflection': False, 'title': 'GitHub - UKPLab/sentence-transformers: State-of-the-Art Text Embeddings', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-classification-of-rss-articles/160986/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 230932, 'name': 'John do', 'username': 'JPFrancoia', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/dbc845/{size}.png', 'created_at': '2025-07-03T18:07:33.404Z', 'cooked': 'Hi,
\nThank you for your answer and sorry for the late reply (got distracted by work, life, etc).
\nI have read/watched some of the resources you sent (this video in particular is really nice: https://www.youtube.com/watch?v=uk6SlTzfbUY) and I now have a basic grasp of how positive unlabelled learning works.
I have implemented two approaches with the following algorithms:
\nSince last time, I built a very modest dataset of “bad” articles: articles I don’t want to read, I don’t find them interesting. I have labelled 70 of them, I intend to use them in my validation set.
\nMy approach is:
\n(# of correct good + # of correct bad) / (total good + total bad)During validation:
\nSame for bad.
\nMy approach is:
\n(# of correct good + # of correct bad) / (total good + total bad)I got insane results and they feel too good to be true:
\nNOTE: I have done a bit of parameter tuning on the OneClassSVM but not on the WeightedElkanotoPuClassifier.
\nimport asyncio\n\nimport numpy as np\nfrom bs4 import BeautifulSoup\nfrom cleantext import clean\nfrom sentence_transformers import SentenceTransformer\n# from sklearn.model_selection import GridSearchCV\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.svm import OneClassSVM\n\nfrom feedoscope.data_registry import data_registry as dr\n\nMODEL_NAME = ""sentence-transformers/all-MiniLM-L12-v2""\n\n\ndef strip_html_keep_text(html: str) -> str:\n soup = BeautifulSoup(html, ""html.parser"")\n text = soup.get_text(separator="" "", strip=True)\n return "" "".join(text.split())\n\n\ndef compute_embeddings(model, texts: list[str]):\n embeddings = model.encode(\n texts, show_progress_bar=True, normalize_embeddings=True, convert_to_numpy=True\n )\n return embeddings\n\n\ndef prepare_articles_text(articles) -> list[str]:\n texts = []\n for a in articles:\n text = clean(\n strip_html_keep_text(f""{a[\'feed_name\']} {a[\'title\']} {a[\'content\']}"")\n )\n texts.append(text)\n\n return texts\n\n\ndef normalize_scores(scores):\n scaler = MinMaxScaler()\n return scaler.fit_transform(scores.reshape(-1, 1)).flatten()\n\n\ndef ocsvm_score(estimator, X):\n # Higher decision_function means more inlier-like\n return np.mean(estimator.decision_function(X))\n\n\nasync def main() -> None:\n print(""Loading SentenceTransformer model..."")\n model = SentenceTransformer(MODEL_NAME)\n print(""Model loaded successfully."")\n\n print(""Collecting articles from the database..."")\n await dr.global_pool.open(wait=True)\n articles = await dr.get_articles()\n print(f""Collected {len(articles)} articles."")\n\n print(""Computing embeddings for articles..."")\n embeddings = compute_embeddings(model, prepare_articles_text(articles))\n print(f""Computed embeddings for {len(embeddings)} articles."")\n\n # Use best parameters directly\n ocsvm = OneClassSVM(kernel=""linear"", gamma=""scale"", nu=0.2)\n ocsvm.fit(embeddings)\n\n # # Hyperparameter tuning for OneClassSVM\n # param_grid = {\n # ""kernel"": [""rbf"", ""linear"", ""sigmoid""],\n # ""gamma"": [""scale"", ""auto"", 0.01, 0.1, 1],\n # ""nu"": [0.01, 0.05, 0.1, 0.2]\n # }\n # print(""Tuning OneClassSVM hyperparameters..."")\n # ocsvm = OneClassSVM()\n # grid = GridSearchCV(\n # OneClassSVM(),\n # param_grid,\n # cv=3,\n # n_jobs=-1,\n # scoring=ocsvm_score\n # )\n # grid.fit(embeddings)\n # best_ocsvm = grid.best_estimator_\n # print(""Best parameters:"", grid.best_params_)\n\n not_good_sample = await dr.get_sample_not_good()\n not_good_embeddings = compute_embeddings(\n model, prepare_articles_text(not_good_sample)\n )\n raw_scores = ocsvm.decision_function(not_good_embeddings)\n scores = normalize_scores(raw_scores)\n\n correct_not_good, total_good = sum(s <= 0.5 for s in scores), len(scores)\n\n good_sample = await dr.get_sample_good()\n good_embeddings = compute_embeddings(model, prepare_articles_text(good_sample))\n raw_scores = ocsvm.decision_function(good_embeddings)\n scores = normalize_scores(raw_scores)\n\n correct_good, total_not_good = sum(s > 0.5 for s in scores), len(scores)\n\n print(\n f""Overall precision: {(correct_good + correct_not_good) / (total_good + total_not_good):.2f}""\n )\n\n\nif __name__ == ""__main__"":\n asyncio.run(main())\nimport asyncio\n\nimport numpy as np\nfrom bs4 import BeautifulSoup\nfrom cleantext import clean\nfrom pulearn import WeightedElkanotoPuClassifier\nfrom sentence_transformers import SentenceTransformer\nfrom sklearn.svm import SVC\n\nfrom feedoscope.data_registry import data_registry as dr\n\nMODEL_NAME = ""sentence-transformers/all-MiniLM-L12-v2""\n\n\ndef strip_html_keep_text(html: str) -> str:\n soup = BeautifulSoup(html, ""html.parser"")\n text = soup.get_text(separator="" "", strip=True)\n return "" "".join(text.split())\n\n\ndef compute_embeddings(model, texts: list[str]):\n embeddings = model.encode(\n texts, show_progress_bar=True, normalize_embeddings=True, convert_to_numpy=True\n )\n return embeddings\n\n\ndef prepare_articles_text(articles) -> list[str]:\n texts = []\n for a in articles:\n text = clean(\n strip_html_keep_text(f""{a[\'feed_name\']} {a[\'title\']} {a[\'content\']}"")\n )\n texts.append(text)\n\n return texts\n\n\nasync def main() -> None:\n\n print(""Loading SentenceTransformer model..."")\n model = SentenceTransformer(MODEL_NAME)\n print(""Model loaded successfully."")\n\n print(""Collecting articles from the database..."")\n await dr.global_pool.open(wait=True)\n articles = await dr.get_articles()\n print(f""Collected {len(articles)} articles."")\n\n print(""Computing embeddings for articles..."")\n embeddings = compute_embeddings(model, prepare_articles_text(articles))\n print(f""Computed embeddings for {len(embeddings)} articles."")\n\n print(""Collecting unread articles from the database..."")\n await dr.global_pool.open(wait=True)\n unlabeled_articles = await dr.get_unread_articles()\n print(f""Collected {len(unlabeled_articles)} unread articles."")\n\n print(""Computing embeddings for unread articles..."")\n unlabeled_embeddings = compute_embeddings(\n model, prepare_articles_text(unlabeled_articles)\n )\n print(f""Computed embeddings for {len(unlabeled_embeddings)} unread articles."")\n\n # Combine embeddings and labels for PU learning\n X = np.concatenate([embeddings, unlabeled_embeddings], axis=0)\n y = np.concatenate(\n [np.ones(len(embeddings)), np.zeros(len(unlabeled_embeddings))], axis=0\n )\n\n print(""Fitting PU classifier..."")\n\n # Takes a while for 7k + 7k articles\n svc = SVC(C=10, kernel=""rbf"", gamma=0.4, probability=True)\n\n # svc = SVC(C=10, kernel=\'linear\', gamma=\'scale\', probability=True)\n\n pu_estimator = WeightedElkanotoPuClassifier(\n estimator=svc,\n labeled=len(embeddings),\n unlabeled=len(unlabeled_embeddings),\n hold_out_ratio=0.2,\n )\n pu_estimator.fit(X, y)\n\n print(""PU classifier fitted successfully."")\n\n not_good_sample = await dr.get_sample_not_good()\n not_good_embeddings = compute_embeddings(\n model, prepare_articles_text(not_good_sample)\n )\n scores = pu_estimator.predict_proba(not_good_embeddings)[:, 1]\n\n correct_not_good, total_good = sum(s <= 0.5 for s in scores), len(scores)\n\n good_sample = await dr.get_sample_good()\n good_embeddings = compute_embeddings(model, prepare_articles_text(good_sample))\n scores = pu_estimator.predict_proba(good_embeddings)[:, 1]\n\n correct_good, total_not_good = sum(s > 0.5 for s in scores), len(scores)\n\n print(\n f""Overall precision: {(correct_good + correct_not_good) / (total_good + total_not_good):.2f}""\n )\n\n breakpoint()\n\n\nif __name__ == ""__main__"":\n asyncio.run(main())\n\nThere does not seem to be any particular problem, but if the figures are too good, data leakage may be suspected.
\n', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-07-04T00:34:24.590Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 160986, 'topic_slug': 'text-classification-of-rss-articles', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://www.geeksforgeeks.org/machine-learning/what-is-data-leakage/', 'internal': False, 'reflection': False, 'title': 'What is Data Leakage? - GeeksforGeeks', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-classification-of-rss-articles/160986/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 231099, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-04T21:20:55.581Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-07-04T21:20:55.581Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 160986, 'topic_slug': 'text-classification-of-rss-articles', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/text-classification-of-rss-articles/160986/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello!
+I’m a software engineer with good coding skills but limited knowledge about AI. I have embarked in a simple project.
+I have a large amount of RSS articles that I have read or liked. I consider these “interesting”. I then have about a gazillion unread articles. These can be interesting, but are most likely uninteresting since I haven’t read them.
+My goal is, for any new article, to compute a score of interesting-ness. This will help me quickly identify the articles worth reading.
The articles range in length from 400 to 4000 tokens. I have about 5000 read/liked articles. I was tempted to take about 5000 unread articles, label them as not_important, take all my liked/read articles and label them as important. Then train a binary classifier. Something like what is described in the hugging face website: Text classification. I used distilbert/distilbert-base-uncased like in the tutorial, and followed almost exactly the steps of the tutorial.
{'loss': 0.6051, 'grad_norm': 2.22690749168396, 'learning_rate': 6.162420382165605e-06, 'epoch': 1.59}
+{'eval_loss': 0.5926874279975891, 'eval_accuracy': 0.6693258875149581, 'eval_runtime': 357.0262, 'eval_samples_per_second': 7.022, 'eval_steps_per_second': 0.221, 'epoch': 2.0}
+{'train_runtime': 12047.1712, 'train_samples_per_second': 1.665, 'train_steps_per_second': 0.052, 'train_loss': 0.592256072220529, 'epoch': 2.0}
+I got modest results after training.
+The question I have for this forum is this one: is it the right approach and should I persevere? Should I put some effort into trying to get a better dataset (like trying to label my not_important articles better), or is there a better approach?
+For example, I have also considered using the model to calculate the embeddings of all the read/liked articles and using a “traditional” algorithm like SVM to train a one class classifier, instead of a binary one.
+The bottleneck to improving the accuracy of the model will be to properly label “not_important” article, if there was a way to get away with not doing that, that would be great
Please let me know what you think
","Hello.
+Given that it works reasonably well in practice, I think the approach is correct. There are many successor models to BERT, so it should be possible to improve accuracy using those.
+Another approach that can be taken when there is little labeled data is something called Positive Unlabeled Learning…
+Another common approach is to use commercial AI to create a training dataset using your own data. This is almost always effective if the budget allows. However, in this case, there is already a considerable amount of data available, so it may be sufficient to process the data using Python.
+Resources:
+ +" +No (0) models returned by ‘Text2Text’ search filter,https://discuss.huggingface.co/t/no-0-models-returned-by-text2text-search-filter/161546,161546,2,2025-07-02 15:36:06.503000+00:00,"[{'id': 230709, 'name': 'Dom', 'username': 'Substance', 'avatar_template': '/user_avatar/discuss.huggingface.co/substance/{size}/50494_2.png', 'created_at': '2025-07-02T15:36:06.565Z', 'cooked': 'Hello,
\nMy colleague reported to me that the ‘Text2Text’ search filter returned 0 models (it was working for them earlier today). I’ve also tested it out myself, and it intermittently returns some model results (sometimes it does show models, but most of the time, it shows no models).
\nWe’ve tried hard-refreshing both our browsers and trying in separate tabs/browsers, but it doesn’t seem to help. All other search filters work fine.
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-02T15:36:06.565Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 14, 'readers_count': 13, 'score': 92.8, 'yours': False, 'topic_id': 161546, 'topic_slug': 'no-0-models-returned-by-text2text-search-filter', 'display_username': 'Dom', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98488, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/no-0-models-returned-by-text2text-search-filter/161546/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 230711, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-02T15:42:28.523Z', 'cooked': 'I don’t really understand the background, but everyone is in that situation right now.
\n\nI’m not sure if this is related to Hugging Chat ending…
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-02T15:42:28.523Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 13, 'readers_count': 12, 'score': 17.6, 'yours': False, 'topic_id': 161546, 'topic_slug': 'no-0-models-returned-by-text2text-search-filter', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/huggingchat/chat-ui/discussions/747', 'internal': False, 'reflection': False, 'title': 'huggingchat/chat-ui · [ANNOUNCEMENT] 📣 HuggingChat is closing for now', 'clicks': 9}, {'url': 'https://discuss.huggingface.co/t/are-inferenceclient-s-down/161485/4', 'internal': True, 'reflection': False, 'title': ""Are InferenceClient()'s down?"", 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/no-0-models-returned-by-text2text-search-filter/161546/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230842, 'name': 'Lucain Pouget', 'username': 'Wauplin', 'avatar_template': '/user_avatar/discuss.huggingface.co/wauplin/{size}/40815_2.png', 'created_at': '2025-07-03T08:27:19.271Z', 'cooked': 'Hi there, all “text2text-generation” models have been moved to “text-generation”. Semantically these 2 tags are not exactly the same but having both was quite confusing to a lot of users. We preferred merging both in the bigger category “text-generation”.
\n(we need to remove the “text2text-generation” filter though)
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-03T08:27:19.271Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 10, 'readers_count': 9, 'score': 52.0, 'yours': False, 'topic_id': 161546, 'topic_slug': 'no-0-models-returned-by-text2text-search-filter', 'display_username': 'Lucain Pouget', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 9207, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/no-0-models-returned-by-text2text-search-filter/161546/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 230944, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-03T20:27:22.892Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-03T20:27:22.892Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 161546, 'topic_slug': 'no-0-models-returned-by-text2text-search-filter', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/no-0-models-returned-by-text2text-search-filter/161546/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+My colleague reported to me that the ‘Text2Text’ search filter returned 0 models (it was working for them earlier today). I’ve also tested it out myself, and it intermittently returns some model results (sometimes it does show models, but most of the time, it shows no models).
+We’ve tried hard-refreshing both our browsers and trying in separate tabs/browsers, but it doesn’t seem to help. All other search filters work fine.
+","Hi there, all “text2text-generation” models have been moved to “text-generation”. Semantically these 2 tags are not exactly the same but having both was quite confusing to a lot of users. We preferred merging both in the bigger category “text-generation”.
+(we need to remove the “text2text-generation” filter though)
" +Video and picture making ai,https://discuss.huggingface.co/t/video-and-picture-making-ai/161564,161564,5,2025-07-02 17:01:58.199000+00:00,"[{'id': 230736, 'name': 'da jewelz', 'username': 'dajewelz', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/b5ac83/{size}.png', 'created_at': '2025-07-02T17:01:58.257Z', 'cooked': 'hello, I was wondering what would be the best ai for me to download from here, I want an ai model that I can feed my own artwork into it so then I can have help making some short form content with it. I would be making videos from ranges 15 min- 30 min and will be storing this ai model on a Mac. Help is very much appreciated on how to download/use/find the right ai model for me. Thank you for looking at this post, and thank you for commenting
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-02T17:01:58.257Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 108, 'reads': 12, 'readers_count': 11, 'score': 517.4, 'yours': False, 'topic_id': 161564, 'topic_slug': 'video-and-picture-making-ai', 'display_username': 'da jewelz', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 69447, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/video-and-picture-making-ai/161564/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 230737, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-02T17:15:36.662Z', 'cooked': 'Video generation models themselves have become increasingly available as open source, but generating long videos requires considerable computing power…
\nThe quickest way to find a promising model is to check out Spaces.
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-02T17:15:36.662Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 12, 'readers_count': 11, 'score': 42.4, 'yours': False, 'topic_id': 161564, 'topic_slug': 'video-and-picture-making-ai', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B', 'internal': False, 'reflection': False, 'title': 'Wan-AI/Wan2.1-VACE-1.3B · Hugging Face', 'clicks': 11}, {'url': 'https://huggingface.co/spaces?category=video-generation&sort=trending', 'internal': False, 'reflection': False, 'title': 'Spaces - Hugging Face', 'clicks': 8}, {'url': 'https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged', 'internal': False, 'reflection': False, 'title': 'Comfy-Org/Wan_2.1_ComfyUI_repackaged · Hugging Face', 'clicks': 5}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/video-and-picture-making-ai/161564/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 230738, 'name': 'da jewelz', 'username': 'dajewelz', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/b5ac83/{size}.png', 'created_at': '2025-07-02T17:27:15.253Z', 'cooked': 'thank you for this information, and thank you for replying
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-07-02T17:27:15.253Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 16.8, 'yours': False, 'topic_id': 161564, 'topic_slug': 'video-and-picture-making-ai', 'display_username': 'da jewelz', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 69447, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/video-and-picture-making-ai/161564/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230913, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-03T14:58:28.321Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-07-03T14:58:28.321Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 161564, 'topic_slug': 'video-and-picture-making-ai', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/video-and-picture-making-ai/161564/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","hello, I was wondering what would be the best ai for me to download from here, I want an ai model that I can feed my own artwork into it so then I can have help making some short form content with it. I would be making videos from ranges 15 min- 30 min and will be storing this ai model on a Mac. Help is very much appreciated on how to download/use/find the right ai model for me. Thank you for looking at this post, and thank you for commenting
","Video generation models themselves have become increasingly available as open source, but generating long videos requires considerable computing power…
+The quickest way to find a promising model is to check out Spaces.
+ +" +Spaces category filters,https://discuss.huggingface.co/t/spaces-category-filters/161550,161550,24,2025-07-02 15:50:29.928000+00:00,"[{'id': 230715, 'name': 'Anthony Noto', 'username': 'thankfulcarp', 'avatar_template': '/user_avatar/discuss.huggingface.co/thankfulcarp/{size}/50499_2.png', 'created_at': '2025-07-02T15:50:30.010Z', 'cooked': 'I recently made a space I am pretty proud of using the latest fusionx wan model and 29 different loras. It does image to video but does not show up in the image to video filter on spaces hub. How do I set the category filter so people can find my project?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-02T15:50:30.010Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 8, 'readers_count': 7, 'score': 56.6, 'yours': False, 'topic_id': 161550, 'topic_slug': 'spaces-category-filters', 'display_username': 'Anthony Noto', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/thankfulcarp/Wan_FusionX_with_Loras', 'internal': False, 'reflection': False, 'title': 'Wan I2V FusionX With Loras - a Hugging Face Space by thankfulcarp', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98491, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-category-filters/161550/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 230721, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-02T16:04:40.685Z', 'cooked': 'Since there are no items where the space creator explicitly sets categories, I think categories are probably automatically generated by AI. I think title and short_description are used as judgment criteria by AI, so it might be better to specify them explicitly.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-07-02T16:04:40.685Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 16.6, 'yours': False, 'topic_id': 161550, 'topic_slug': 'spaces-category-filters', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/hub/spaces-config-reference', 'internal': False, 'reflection': False, 'title': 'Spaces Configuration Reference', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/spaces-category-filters/161550/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 230802, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-03T04:04:50.049Z', 'cooked': '\n
short_description: string A short description of the Space. This will be displayed in the Space’s thumbnail.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-07-03T04:04:50.049Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 161550, 'topic_slug': 'spaces-category-filters', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/spaces-category-filters/161550/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I recently made a space I am pretty proud of using the latest fusionx wan model and 29 different loras. It does image to video but does not show up in the image to video filter on spaces hub. How do I set the category filter so people can find my project?
","Since there are no items where the space creator explicitly sets categories, I think categories are probably automatically generated by AI. I think title and short_description are used as judgment criteria by AI, so it might be better to specify them explicitly.
+" +Using datasets to open jsonl,https://discuss.huggingface.co/t/using-datasets-to-open-jsonl/161037,161037,10,2025-06-28 18:33:58.353000+00:00,"[{'id': 229909, 'name': 'bluebingo', 'username': 'bluebingo', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/f4b2a3/{size}.png', 'created_at': '2025-06-28T18:33:58.407Z', 'cooked': '+
short_description: string A short description of the Space. This will be displayed in the Space’s thumbnail.
I am trying to open a JSONL format file using the datasets library. Here is my code:
from datasets import load_dataset\n\npath = ""./testdata.jsonl""\ndataset = load_dataset(\'json\', data_files=path, split=\'train\')\nThe contents of testdata.jsonl are organized as follows (just for testing):
\n{""src"":""hello"",""term"":{""a"":""aa""}}\n{""src"":""hi"",""term"":{""b"":""bb""}}\nWhen I use the code above to load the dataset and attempt to print the second item, like this:
\nprint(dataset[1])\nI get the following output:
\n{\'src\': \'hi\', \'term\': {\'a\': None, \'b\': \'bb\'}}\nInstead of the expected output:
\n{\'src\': \'hi\', \'term\': {\'b\': \'bb\'}}\nHow can I obtain the second format of the dataset? Is it possible that I simply forgot to include a parameter?
', 'post_number': 1, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-06-28T18:56:54.940Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 46, 'reads': 8, 'readers_count': 7, 'score': 246.6, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'bluebingo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98155, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 229932, 'name': 'Andrew Scott', 'username': 'Pimpcat-AU', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png', 'created_at': '2025-06-28T22:47:45.598Z', 'cooked': 'Ensure the JSONL file is correctly formatted:
\nEach line in the file should be a valid JSON object with no extra commas or brackets. For example, the file should look like this:
{“src”:“hello”,“term”:{“a”:“aa”}}
\n{“src”:“hi”,“term”:{“b”:“bb”}}
After fixing the JSONL format, use the following code to load the dataset properly:
\nfrom datasets import load_dataset
\npath = “./testdata.jsonl”
\ndataset = load_dataset(‘json’, data_files=path, split=‘train’)
print(dataset[1]) # This should now work correctly
\nAfter these changes, the second entry should now print the correct data:
\n{‘src’: ‘hi’, ‘term’: {‘b’: ‘bb’}}
\nAlso, ensure there are no extra spaces or line breaks in the dataset if it’s large. Each line should be a valid JSON object.
\nResponse generated by Triskel Data Deterministic Ai
', 'post_number': 2, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-06-28T22:48:34.808Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 21.4, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'Andrew Scott', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96276, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229934, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-28T22:55:56.602Z', 'cooked': 'Another option, albeit a bit rough, is this:
\nfrom datasets import load_dataset\n\ndef process(example):\n example[""term""] = str({k: v for k, v in example[""term""].items() if v is not None})\n return example\n\npath = ""./testdata.jsonl""\ndataset = load_dataset(\'json\', data_files=path, split=\'train\')\n\nprint(dataset[1]) # {\'src\': \'hi\', \'term\': {\'a\': None, \'b\': \'bb\'}}\n\ndataset = dataset.map(process)\n\nprint(dataset[1]) # {\'src\': \'hi\', \'term\': ""{\'b\': \'bb\'}""}\nThank you for your advice. I appreciate your efforts, but unfortunately, it hasn’t been effective for me.
', 'post_number': 4, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-06-29T18:35:49.044Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'bluebingo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 96276, 'username': 'Pimpcat-AU', 'name': 'Andrew Scott', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98155, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230035, 'name': 'bluebingo', 'username': 'bluebingo', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/f4b2a3/{size}.png', 'created_at': '2025-06-29T18:38:28.361Z', 'cooked': 'Thank you for your advice; it was really helpful in solving the problem! However, I find it a bit cumbersome to map the datasets each time I want to open a JSONL file with JSON elements. I wonder if there might be a more permanent solution to address this issue.
', 'post_number': 5, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-06-29T18:38:28.361Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'bluebingo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98155, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230064, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-30T01:50:35.067Z', 'cooked': '\n\nI find it a bit cumbersome to map the datasets each time I want to open a JSONL file with JSON elements. I wonder if there might be a more permanent solution to address this issue.
\n
That’s true. There may be a more concise method (including potential ones). I’ll mention it to the library developer. @lhoestq
', 'post_number': 8, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-06-30T01:50:35.067Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 6, 'readers_count': 5, 'score': 31.2, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/8', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230094, 'name': 'bluebingo', 'username': 'bluebingo', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/f4b2a3/{size}.png', 'created_at': '2025-06-30T08:03:11.121Z', 'cooked': 'Thank you! I look forward to any official solutions that the developer might provide.
', 'post_number': 9, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-06-30T08:03:11.121Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'bluebingo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98155, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230360, 'name': 'Quentin Lhoest', 'username': 'lhoestq', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png', 'created_at': '2025-07-01T12:27:46.538Z', 'cooked': 'Hi ! This behavior is expected since datasets uses Arrow which has fixed types. This means each sample should have the same subfields with the same types. Missing subfields are filled with None.
You can restructure your data differently to fit this paradigm: either converting nested data as one string, or use one list for keys and one list for values.
', 'post_number': 10, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-07-01T12:27:46.538Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 28, 'reads': 6, 'readers_count': 5, 'score': 171.2, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'Quentin Lhoest', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 76, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/10', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 230443, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-07-01T20:18:09.947Z', 'cooked': 'Thank you, lhonestq!
', 'post_number': 11, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-07-01T20:18:09.947Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/11', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230493, 'name': 'bluebingo', 'username': 'bluebingo', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/f4b2a3/{size}.png', 'created_at': '2025-07-02T01:16:11.203Z', 'cooked': 'Thank you, lhonestq!
', 'post_number': 12, 'post_type': 1, 'posts_count': 11, 'updated_at': '2025-07-02T01:16:11.203Z', 'reply_count': 0, 'reply_to_post_number': 10, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'bluebingo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76, 'username': 'lhoestq', 'name': 'Quentin Lhoest', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98155, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/using-datasets-to-open-jsonl/161037/12', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230678, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-07-02T13:17:03.260Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 13, 'post_type': 3, 'posts_count': 11, 'updated_at': '2025-07-02T13:17:03.260Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 161037, 'topic_slug': 'using-datasets-to-open-jsonl', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/using-datasets-to-open-jsonl/161037/13', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to open a JSONL format file using the datasets library. Here is my code:
from datasets import load_dataset
+
+path = ""./testdata.jsonl""
+dataset = load_dataset('json', data_files=path, split='train')
+The contents of testdata.jsonl are organized as follows (just for testing):
+{""src"":""hello"",""term"":{""a"":""aa""}}
+{""src"":""hi"",""term"":{""b"":""bb""}}
+When I use the code above to load the dataset and attempt to print the second item, like this:
+print(dataset[1])
+I get the following output:
+{'src': 'hi', 'term': {'a': None, 'b': 'bb'}}
+Instead of the expected output:
+{'src': 'hi', 'term': {'b': 'bb'}}
+How can I obtain the second format of the dataset? Is it possible that I simply forgot to include a parameter?
","Thank you, lhonestq!
" +How to upload documents to the SupabaseVectorStore?,https://discuss.huggingface.co/t/how-to-upload-documents-to-the-supabasevectorstore/161245,161245,24,2025-07-01 00:22:19.997000+00:00,"[{'id': 230232, 'name': 'Sen Li', 'username': 'AllIllusion', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/e9c0ed/{size}.png', 'created_at': '2025-07-01T00:22:20.073Z', 'cooked': 'Hi everyone,
\nI am learning RAG for GAIA, from here: test.ipynb · baixianger/RobotPai at main
\nHowever, I was not able to upload documents to Supabase, as shown in screenshots:
\nI have tried two ways:
\n# wrap the metadata.jsonl\'s questions and answers into a list of document\nlistDict_QA_Doc = []\nfor dict_RandomQA in listDict_Metadata:\n strQA_Content = f""Question : {dict_RandomQA[\'Question\']}\\n\\nFinal answer : {dict_RandomQA[\'Final answer\']}""\n dict_QA_Doc = {\n ""id"": dict_RandomQA[\'task_id\'],\n ""content"" : strQA_Content,\n ""metadata"" : {\n ""source"" : dict_RandomQA[\'task_id\']\n },\n ""embedding"" : embeddings.embed_query(strQA_Content),\n }\n listDict_QA_Doc.append(dict_QA_Doc)\n\n\nresponse = syncClient.table(""documents"").insert(listDict_QA_Doc).execute()\nand
\n# wrap the metadata.jsonl\'s questions and answers into a list of document\nlistDoc_QA_Metadata = []\nfor dict_Metadata in listDict_Metadata:\n strQA_Content = f""Question : {dict_Metadata[\'Question\']}\\n\\nFinal answer : {dict_Metadata[\'Final answer\']}""\n doc_QA_Metadata = Document(\n id = dict_Metadata[\'task_id\'],\n page_content = strQA_Content,\n metadata = {""source"": dict_Metadata[\'task_id\']},\n embedding = embeddings.embed_query(strQA_Content)\n )\n listDoc_QA_Metadata.append(doc_QA_Metadata)\n\n\nvector_store = SupabaseVectorStore.from_documents(\n listDoc_QA_Metadata,\n embeddings,\n client=syncClient,\n table_name=""documents"",\n query_name=""match_documents"",\n)\nHowever, always get the same error:
\nError inserting data into Supabase: {\'message\': \'JSON could not be generated\', \'code\': 404, \'hint\': \'Refer to full message for details\', \'details\': ""b\'{}\'""}\nCould anyone please help?
How about changing the version of pydantic?
pip install pydantic==2.10.6\nJust tested, still the same error
Hmm… In that case, could it be that the data you passed is not in the expected JSON structure, as indicated by the error message?
\nYou can verify this by passing extremely simple sample data that is expected to be passed, rather than the actual data.
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-07-01T20:16:20.877Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 161245, 'topic_slug': 'how-to-upload-documents-to-the-supabasevectorstore', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-upload-documents-to-the-supabasevectorstore/161245/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230453, 'name': 'Sen Li', 'username': 'AllIllusion', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/e9c0ed/{size}.png', 'created_at': '2025-07-01T21:23:36.192Z', 'cooked': '\nSolved. Need to create a table on supabase before uploading.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-07-02T12:43:03.536Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 161245, 'topic_slug': 'how-to-upload-documents-to-the-supabasevectorstore', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-upload-documents-to-the-supabasevectorstore/161245/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I am learning RAG for GAIA, from here: test.ipynb · baixianger/RobotPai at main
+However, I was not able to upload documents to Supabase, as shown in screenshots:
+I have tried two ways:
+# wrap the metadata.jsonl's questions and answers into a list of document
+listDict_QA_Doc = []
+for dict_RandomQA in listDict_Metadata:
+ strQA_Content = f""Question : {dict_RandomQA['Question']}\n\nFinal answer : {dict_RandomQA['Final answer']}""
+ dict_QA_Doc = {
+ ""id"": dict_RandomQA['task_id'],
+ ""content"" : strQA_Content,
+ ""metadata"" : {
+ ""source"" : dict_RandomQA['task_id']
+ },
+ ""embedding"" : embeddings.embed_query(strQA_Content),
+ }
+ listDict_QA_Doc.append(dict_QA_Doc)
+
+
+response = syncClient.table(""documents"").insert(listDict_QA_Doc).execute()
+and
+# wrap the metadata.jsonl's questions and answers into a list of document
+listDoc_QA_Metadata = []
+for dict_Metadata in listDict_Metadata:
+ strQA_Content = f""Question : {dict_Metadata['Question']}\n\nFinal answer : {dict_Metadata['Final answer']}""
+ doc_QA_Metadata = Document(
+ id = dict_Metadata['task_id'],
+ page_content = strQA_Content,
+ metadata = {""source"": dict_Metadata['task_id']},
+ embedding = embeddings.embed_query(strQA_Content)
+ )
+ listDoc_QA_Metadata.append(doc_QA_Metadata)
+
+
+vector_store = SupabaseVectorStore.from_documents(
+ listDoc_QA_Metadata,
+ embeddings,
+ client=syncClient,
+ table_name=""documents"",
+ query_name=""match_documents"",
+)
+However, always get the same error:
+Error inserting data into Supabase: {'message': 'JSON could not be generated', 'code': 404, 'hint': 'Refer to full message for details', 'details': ""b'{}'""}
+Could anyone please help?
Solved. Need to create a table on supabase before uploading.
I work inside a secure corporate VPN network, so I’m unable to download Huggingface models using from_pretrained commands. However, I can request the security team to whitelist certain URLs needed for my use-case.
The security team has already whitelisted the ‘huggingface.co’ and ‘cdn-lfs.huggingface.co’ URLs. I can now download the files from repo but the loading functions from_pretrained still don’t work.
I think it’s getting blocked while redirecting the requests internally. So, is there a way to know all (hop) URLs I can request to whitelist to make the load functions work?
\nThanks in advance.
', 'post_number': 1, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-26T14:09:18.971Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9350, 'reads': 117, 'readers_count': 116, 'score': 46513.4, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Ashwani', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://cdn-lfs.huggingface.co', 'internal': False, 'reflection': False, 'clicks': 187}, {'url': 'http://huggingface.co', 'internal': False, 'reflection': False, 'title': 'Hugging Face – The AI community building the future.', 'clicks': 86}, {'url': 'https://discuss.huggingface.co/t/how-to-whitelist-a-hf-space-to-use-brightdata-with-it/143796', 'internal': True, 'reflection': True, 'title': 'How to whitelist a HF space to use brightdata with it?', 'clicks': 11}, {'url': 'https://discuss.huggingface.co/t/cas-bridge-xethub-hf-co-broke/158626/2', 'internal': True, 'reflection': True, 'title': 'Cas-bridge.xethub.hf.co broke', 'clicks': 9}, {'url': 'https://discuss.huggingface.co/t/i-cannot-download-any-large-models-stored-in-xet-with-brave-or-ms-edge-for-weeks/166454/5', 'internal': True, 'reflection': True, 'title': 'I cannot download any large models stored in xet with Brave or MS Edge for weeks', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 10}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 14513, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/1', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 10}], 'current_user_reaction': None, 'reaction_users_count': 10, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 56027, 'name': 'Eliott Coyac', 'username': 'coyotte508', 'avatar_template': '/user_avatar/discuss.huggingface.co/coyotte508/{size}/36751_2.png', 'created_at': '2023-01-26T15:48:50.016Z', 'cooked': 'hi @ayadav
\nCan you give more details, like error logs, etc?
', 'post_number': 2, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-01-26T15:48:50.016Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 114, 'readers_count': 113, 'score': 107.8, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Eliott Coyac', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 6451, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 86846, 'name': 'Brian Law', 'username': 'Data-drone', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/7ea924/{size}.png', 'created_at': '2023-08-30T03:58:37.848Z', 'cooked': 'Is there any update on this?
', 'post_number': 3, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-08-30T03:58:37.848Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 30, 'reads': 93, 'readers_count': 92, 'score': 183.6, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Brian Law', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 5630, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/3', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 95802, 'name': 'Nik Kramaric', 'username': 'cosmo88', 'avatar_template': '/user_avatar/discuss.huggingface.co/cosmo88/{size}/20569_2.png', 'created_at': '2023-10-23T17:34:06.412Z', 'cooked': 'Having the same issue. Is there a listing of URLs that we can whitelist? Also if there are any planned changes to URLs is there a roadmap so we can stay on top of it?
', 'post_number': 4, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-10-23T17:34:06.412Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 28, 'reads': 85, 'readers_count': 84, 'score': 172.0, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Nik Kramaric', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31863, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 99563, 'name': 'kearney', 'username': 'kearney', 'avatar_template': '/user_avatar/discuss.huggingface.co/kearney/{size}/21274_2.png', 'created_at': '2023-11-17T13:50:16.592Z', 'cooked': 'I’ll try to supply error logs next time I encounter it, but it has come up multiple times for me as well. When we try to call <model>.from_pretrained(""repo"") in our DataBricks environment, we get an SSL error about not having the proper certificate. We’ve also gotten a max_retries error but I can’t say for certain if that was due to the underlying whitelist request. There are ways around this, but if HF published a domain list that we could use to properly configure our environments, that would be very useful!
hi! any updates on this? or any alternatives to follow meanwhile? I am about to try downloading a model and going offline and then pushing it up to databricks. Yet, if you had a better idea, or tried this before, I’d like to hear.
', 'post_number': 6, 'post_type': 1, 'posts_count': 20, 'updated_at': '2023-11-28T23:43:05.295Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 127, 'reads': 80, 'readers_count': 79, 'score': 631.0, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': None, 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 34668, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 102928, 'name': 'Jimmy Wang', 'username': 'JimmyWang2023', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/eb8c5e/{size}.png', 'created_at': '2023-12-08T09:13:47.653Z', 'cooked': 'I have same issue with download from different cdn name.
\nAfter our IT team added
\nhttp://huggingface.co/ and
\nhttp://cdn-lfs.huggingface.co/ in whitelist.
For example, it is work for download meta-llama/Llama-2-13b-chat.
\nBut error when the cdn become cdn-lfs-us-1.huggingface.co or other regions.
Update? Same issue here. I’ve gotten around by using my home network to connect to the hf repo and download to my workstation cache. Then I reconnect to VPN into the corporate network and copy from my workstation to the server cache. This is painfully slow.
\nFWIW curl -IL test shows redirection (302 responses) from the repo when I am connected to the corporate network (fails to download). However on my home network there are no redirects (successful download). Is there an issue with general redirection handling?
', 'post_number': 8, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-03-28T19:32:53.049Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 75, 'reads': 70, 'readers_count': 69, 'score': 389.0, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'chuck', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 44983, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 160277, 'name': 'Rishav Dash', 'username': 'RishuD7', 'avatar_template': '/user_avatar/discuss.huggingface.co/rishud7/{size}/32370_2.png', 'created_at': '2024-10-05T12:59:17.106Z', 'cooked': 'Hey was anyone able to find a solution for this?
', 'post_number': 9, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-10-05T12:59:17.106Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 141, 'reads': 54, 'readers_count': 53, 'score': 715.8, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Rishav Dash', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 66383, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 160489, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-10-06T03:28:34.240Z', 'cooked': 'Related:
\n', 'post_number': 10, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-10-06T03:28:34.240Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 417, 'reads': 57, 'readers_count': 56, 'score': 2066.4, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/not-able-to-upload-or-download-custom-datasets/110001/3', 'internal': True, 'reflection': False, 'title': 'Not able to upload or download custom datasets', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/10', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 160814, 'name': 'Pierric Cistac', 'username': 'pierric', 'avatar_template': '/user_avatar/discuss.huggingface.co/pierric/{size}/50750_2.png', 'created_at': '2024-10-07T22:01:26.202Z', 'cooked': 'Note that for security reasons, we recently updated the domain for our CDN; in order to be able to download files you also need to whitelist the following domains:
\nwe have created exception for SSL inspection for FQDN listed by pierric plus these 2 ones:
\n\n\nBut it is still does not work, always same error encountered SSL: CERTIFICATE_VERIFY_FAILED when trying to download sentence-transformers/all-MiniLM-L6-v2
', 'post_number': 12, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-12-12T15:11:06.947Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 39, 'reads': 43, 'readers_count': 42, 'score': 208.6, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Remi Le Marois', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co', 'internal': False, 'reflection': False, 'title': 'Hugging Face – The AI community building the future.', 'clicks': 41}, {'url': 'https://hub-ci.huggingface.co', 'internal': False, 'reflection': False, 'title': 'Hugging Face – The AI community building the future.', 'clicks': 23}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 76764, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/12', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 204973, 'name': 'Sean Morgan', 'username': 'sean-pai', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/c6cbf5/{size}.png', 'created_at': '2025-02-24T14:31:46.249Z', 'cooked': 'Hi @pierric has the above list changed since the XetHub announcement?
\nWhile downloading, I’m seeing a domain of cas-bridge.xethub.hf.co as well. Is this the only additional domain or are there others?
Hey @sean-pai, sorry about that, indeed we recently started migrating some repos from LFS to Xet (checkout this blogpost if you want to learn more about Xet).
\nAs a result (and as you found out), you need to add cas-bridge.xethub.hf.co for the download path (I updated my original reply above). We’ll communicate here when we enable the Xet upload path.
Hi @sean-pai, just a quick follow up, we’ve just released the Xet client which can be used to download these repos using the xet format directly. If you are interested in faster downloads of Xet enabled repos, follow these instructions here.
\nIf you install the client and download the same content, you will also need to add two new endpoints, cas-server.xethub.hf.co and transfer.xethub.hf.co.
Hi @brianronan,
\nThe certificate returned for cas-server, is the cas-bridge certificate.
\n\n\n(.venv) mark@wide:~/prog/b3d-lora-trainer$ openssl s_client -connect cas-server.xethub.hf.co:443 -servername cas-server.xethub.hf.co
\nConnecting to 52.71.209.178
\n
\nCONNECTED(00000003)
\ndepth=2 C=US, O=Amazon, CN=Amazon Root CA 1
\nverify return:1
\ndepth=1 C=US, O=Amazon, CN=Amazon RSA 2048 M03
\nverify return:1
\ndepth=0 CN=cas-bridge.xethub.hf.co
\nverify return:1Certificate chain
\n
\n0 s:CN=cas-bridge.xethub.hf.co
\ni:C=US, O=Amazon, CN=Amazon RSA 2048 M03
\na:PKEY: rsaEncryption, 2048 (bit); sigalg: RSA-SHA256
\nv:NotBefore: Jan 29 00:00:00 2025 GMT; NotAfter: Feb 27 23:59:59 2026 GMT
\n-snip-
And thus I get certificate verify failed when using from_pretrained().
\nmodel_name = ""Qwen/Qwen2.5-Coder-7B""\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name,\n trust_remote_code=True,\n torch_dtype=torch.float16,\n device_map=""auto""\n)\n\n', 'post_number': 16, 'post_type': 1, 'posts_count': 20, 'updated_at': '2025-05-26T17:53:32.272Z', 'reply_count': 1, 'reply_to_post_number': 15, 'quote_count': 0, 'incoming_link_count': 36, 'reads': 13, 'readers_count': 12, 'score': 197.6, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Mark', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://cas-server.xethub.hf.co/reconstruction/cd9b3569e15af48b5338d1f02bf99476542809310dde89f1a4301215b1a8a81d%5C', 'internal': False, 'reflection': False, 'clicks': 5}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 60126, 'username': 'brianronan', 'name': 'Brian Ronan', 'avatar_template': '/user_avatar/discuss.huggingface.co/brianronan/{size}/30065_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 60646, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/16', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 224698, 'name': 'Jared Sulzdorf', 'username': 'jsulz', 'avatar_template': '/user_avatar/discuss.huggingface.co/jsulz/{size}/28279_2.png', 'created_at': '2025-05-29T16:45:58.783Z', 'cooked': '“timestamp”:“2025-05-26T17:43:40.209499Z”,“level”:“WARN”,“fields”:{“message”:“Reqwest(reqwest::Error { kind: Request, url: ""https://cas-server.xethub.hf.co/reconstruction/cd9b3569e15af48b5338d1f02bf99476542809310dde89f1a4301215b1a8a81d\\”, source: hyper_util::client::legacy::Error(Connect, Ssl(Error { code: ErrorCode(1), cause: Some(Ssl(ErrorStack([Error { code: 167772294, library: ""SSL routines"", function: ""tls_post_process_server_certificate"", reason: ""certificate verify failed"", file: ""ssl/statem/statem_clnt.c"", line: 2092 }]))) }, X509VerifyResult { code: 20, error: ""unable to get local issuer certificate"" })) }). Retrying…“},“filename”:”/home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs"",“line_number”:175}
\n
Just noting for the followers of this thread that the issue raised here by @marked23 is being handled over here - Certificate Verify Failed cas-server vs. cas-bridge · Issue #351 · huggingface/xet-core · GitHub - and currently seems unrelated to any issues around whitelisting domains.
', 'post_number': 17, 'post_type': 1, 'posts_count': 20, 'updated_at': '2025-05-29T16:45:58.783Z', 'reply_count': 0, 'reply_to_post_number': 16, 'quote_count': 0, 'incoming_link_count': 14, 'reads': 11, 'readers_count': 10, 'score': 87.2, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Jared Sulzdorf', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/xet-core/issues/351', 'internal': False, 'reflection': False, 'title': 'Certificate Verify Failed cas-server vs. cas-bridge · Issue #351 · huggingface/xet-core · GitHub', 'clicks': 61}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 60646, 'username': 'marked23', 'name': 'Mark', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/e95f7d/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 54269, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/17', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230377, 'name': 'Mario Vela', 'username': 'mariovela', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/ed8c4c/{size}.png', 'created_at': '2025-07-01T14:08:50.609Z', 'cooked': 'This was working for us but recently started failing with timeouts whenever we use huggingface_hub (via python or CLI).
\nI noticed we can still download using curl -L https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/model.safetensors?download=true --output model.safetensors but we cannot using
from sentence_transformers import SentenceTransformer\nmodel = SentenceTransformer(\'all-MiniLM-L6-v2\')\nNor using
\nhuggingface-cli download sentence-transformers/all-MiniLM-L6-v2\nBoth of these just hang like:
\nhuggingface-cli download sentence-transformers/all-MiniLM-L6-v2 --max-workers 1\nFetching 30 files: 0%| | 0/30 [00:00<?, ?it/s]Downloading \'model.safetensors\' to \'/home/jupyter/.cache/huggingface/hub/models--sentence-transformers--all-MiniLM-L6-v2/blobs/53aa51172d142c89d9012cce15ae4d6cc0ca6895895114379cacb4fab128d9db.incomplete\'\n\nmodel.safetensors: 0%| | 0.00/90.9M [00:00<?, ?B/s]\n""timestamp"":""2025-07-01T13:40:33.080005Z"",""level"":""WARN"",""fields"":{""message"":""Reqwest(reqwest::Error { kind: Request, url: \\""https://cas-server.xethub.hf.co/reconstruction/789fdf16a3e59f4fbfb6002967ecee539a198dadb5be74ca549aa7dc9b1b55fb\\"", source: hyper_util::client::legacy::Error(Connect, ConnectError(\\""tcp connect error\\"", Os { code: 110, kind: TimedOut, message: \\""Connection timed out\\"" })) }). Retrying...""},""filename"":""/home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs"",""line_number"":200}\n{""timestamp"":""2025-07-01T13:40:33.080067Z"",""level"":""WARN"",""fields"":{""message"":""Retry attempt #0. Sleeping 2.851275886s before the next attempt""},""filename"":""/root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/reqwest-retry-0.7.0/src/middleware.rs"",""line_number"":171}\n{""timestamp"":""2025-07-01T13:58:03.703922Z"",""level"":""WARN"",""fields"":{""message"":""Reqwest(reqwest::Error { kind: Request, url: \\""https://cas-server.xethub.hf.co/reconstruction/789fdf16a3e59f4fbfb6002967ecee539a198dadb5be74ca549aa7dc9b1b55fb\\"", source: hyper_util::client::legacy::Error(Connect, ConnectError(\\""tcp connect error\\"", Os { code: 110, kind: TimedOut, message: \\""Connection timed out\\"" })) }). Retrying...""},""filename"":""/home/runner/work/xet-core/xet-core/cas_client/src/http_client.rs"",""line_number"":200}\n{""timestamp"":""2025-07-01T13:58:03.703998Z"",""level"":""WARN"",""fields"":{""message"":""Retry attempt #1. Sleeping 2.339135315s before the next attempt""},""filename"":""/root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/reqwest-retry-0.7.0/src/middleware.rs"",""line_number"":171}\nIt just hangs and times out for the model.safetensors file.
We have allowlisted:
\ncdn-lfs-us-1.hf.co\ncdn-lfs-eu-1.hf.co\ncdn-lfs.hf.co\ncas-bridge.xethub.hf.co\nAny ideas?
\nIt seems to be going to a cloudfront IP at some point, but I do not know what for and if it is something that can be stopped.
Hi @mariovela
\nCould you try allowlisting the following URLs in addition to the current domains you’ve allowlisted:
\ntransfer.xethub.hf.co\ncas-server.xethub.hf.co\nBoth are used when downloading from/uploading to Xet-enabled repositories when hf-xet is installed.
See @brianronan’s comment above
', 'post_number': 19, 'post_type': 1, 'posts_count': 20, 'updated_at': '2025-07-01T15:15:41.358Z', 'reply_count': 1, 'reply_to_post_number': 18, 'quote_count': 0, 'incoming_link_count': 23, 'reads': 9, 'readers_count': 8, 'score': 136.8, 'yours': False, 'topic_id': 30486, 'topic_slug': 'how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist', 'display_username': 'Jared Sulzdorf', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 98369, 'username': 'mariovela', 'name': 'Mario Vela', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/ed8c4c/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 54269, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-get-a-list-of-all-huggingface-download-redirections-to-whitelist/30486/19', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230384, 'name': 'Mario Vela', 'username': 'mariovela', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/ed8c4c/{size}.png', 'created_at': '2025-07-01T15:18:30.779Z', 'cooked': 'My bad! That works! Thank you!
I work inside a secure corporate VPN network, so I’m unable to download Huggingface models using from_pretrained commands. However, I can request the security team to whitelist certain URLs needed for my use-case.
The security team has already whitelisted the ‘huggingface.co’ and ‘cdn-lfs.huggingface.co’ URLs. I can now download the files from repo but the loading functions from_pretrained still don’t work.
I think it’s getting blocked while redirecting the requests internally. So, is there a way to know all (hop) URLs I can request to whitelist to make the load functions work?
+Thanks in advance.
","Note that for security reasons, we recently updated the domain for our CDN; in order to be able to download files you also need to whitelist the following domains:
+I tried the smolagents WebSearchTool to search some information, but it returns irrelevant information, I don’t know if there is a way to fine-tune the result or the query, attached is the code generated from smolagents and the result
\n
The content seems strange, or rather, it looks like the query isn’t being passed…
\nThere are several implementations of search tools, but if it’s only happening with one of them, the search engine specifications may have changed and the library isn’t compatible.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-28T21:36:53.903Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 161008, 'topic_slug': 'smolagents-websearchtool-search-for-wrong-query', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/smolagents/issues/1386', 'internal': False, 'reflection': False, 'title': 'WebSearchTool example from Guide Tour does not work · Issue #1386 · huggingface/smolagents · GitHub', 'clicks': 7}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/smolagents-websearchtool-search-for-wrong-query/161008/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 230108, 'name': 'doradoradorayaki', 'username': 'dorayaki78', 'avatar_template': '/user_avatar/discuss.huggingface.co/dorayaki78/{size}/50008_2.png', 'created_at': '2025-06-30T10:03:47.381Z', 'cooked': 'Hi the problem is resolved, thanks for your response, it seems that the SSL or TLS handshake doesn’t work properly, and I tried to go to the duckduckgo website and it returns error. But now it is solved, the problem maybe lies in the date and time of the system which is still not in sync with my local time (as I am currently in a different time zone). The other approach is maybe to clear the SSL state
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-30T10:03:47.381Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 20.4, 'yours': False, 'topic_id': 161008, 'topic_slug': 'smolagents-websearchtool-search-for-wrong-query', 'display_username': 'doradoradorayaki', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97781, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/smolagents-websearchtool-search-for-wrong-query/161008/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 230222, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-30T22:04:16.186Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-30T22:04:16.186Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 1, 'readers_count': 0, 'score': 25.2, 'yours': False, 'topic_id': 161008, 'topic_slug': 'smolagents-websearchtool-search-for-wrong-query', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/smolagents-websearchtool-search-for-wrong-query/161008/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I tried the smolagents WebSearchTool to search some information, but it returns irrelevant information, I don’t know if there is a way to fine-tune the result or the query, attached is the code generated from smolagents and the result
+
Hi the problem is resolved, thanks for your response, it seems that the SSL or TLS handshake doesn’t work properly, and I tried to go to the duckduckgo website and it returns error. But now it is solved, the problem maybe lies in the date and time of the system which is still not in sync with my local time (as I am currently in a different time zone). The other approach is maybe to clear the SSL state
" +Text-to-Sql model keeps missing “<” token,https://discuss.huggingface.co/t/text-to-sql-model-keeps-missing-token/158903,158903,6,2025-06-11 11:05:53.474000+00:00,"[{'id': 226936, 'name': 'Brian Antao', 'username': 'BrianAntao', 'avatar_template': '/user_avatar/discuss.huggingface.co/brianantao/{size}/49245_2.png', 'created_at': '2025-06-11T11:05:53.535Z', 'cooked': 'Hello all,
\nI trained the T5-base model using gretelai/synthetic_text_to_sql data set and then fine tuned it on my specific table schema and set of example queries.
\nWhen I test the fine-tuned model it keeps missing the “<” token in the generated query results.
\nI have played with various fine-tuning params – like number of epochs.
\nWhy thus the resultant model not know to use the “<” token ?
\nI added a couple of SQL examples with explicit “<” to the dataset but when I query back it gives me the resulting SQL without the “<” in it which is an incorrect SQL!
\nCheers.
You may need to fine tune the system prompt or validate the generations afterwards with a judge.
\nLeave a like if this helps at all.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-16T08:35:02.767Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 1.2, 'yours': False, 'topic_id': 158903, 'topic_slug': 'text-to-sql-model-keeps-missing-token', 'display_username': 'Riley Fox', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94214, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/text-to-sql-model-keeps-missing-token/158903/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226947, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-11T11:36:53.055Z', 'cooked': 'Hmm… Perhaps tokenizer vocab issue?
\nhttps://stackoverflow.com/questions/75851029/t5-fine-tuned-model-outputs-unk-instead-of-curly-braces-and-other-special-char
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-29T15:39:57.071Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 158903, 'topic_slug': 'text-to-sql-model-keeps-missing-token', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/text-to-sql-model-keeps-missing-token/158903/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello all,
+I trained the T5-base model using gretelai/synthetic_text_to_sql data set and then fine tuned it on my specific table schema and set of example queries.
+When I test the fine-tuned model it keeps missing the “<” token in the generated query results.
+I have played with various fine-tuning params – like number of epochs.
+Why thus the resultant model not know to use the “<” token ?
+I added a couple of SQL examples with explicit “<” to the dataset but when I query back it gives me the resulting SQL without the “<” in it which is an incorrect SQL!
+Cheers.
Hmm… Perhaps tokenizer vocab issue?
+https://stackoverflow.com/questions/75851029/t5-fine-tuned-model-outputs-unk-instead-of-curly-braces-and-other-special-char
Hi I tried to use WebSearchTool from smolagents and got this kind of error, I’m using ollama with model qwen2.5 7b, can anyone help me
\nCode execution failed at line ‘music_recommendations = web_search(query=“best party music”)’ due to: SSLError:
\nHTTPSConnectionPool(host=‘lite.duckduckgo.com’, port=443): Max retries exceeded with url: /lite/?q=best+party+music
\n(Caused by SSLError(SSLCertVerificationError(1, ‘[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed:
\nself-signed certificate (_ssl.c:1028)’)))
I think this might be an SSL error caused by a proxy, VPN, cloud, or internal network firewall, but it’s in the library code…
\nIt might be difficult to work around.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-06-24T13:45:17.856Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 7, 'readers_count': 6, 'score': 11.4, 'yours': False, 'topic_id': 160510, 'topic_slug': 'websearchtool-error', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://stackoverflow.com/questions/51925384/unable-to-get-local-issuer-certificate-when-using-requests', 'internal': False, 'reflection': False, 'clicks': 3}, {'url': 'https://huggingface.co/docs/smolagents/reference/tools#smolagents.WebSearchTool', 'internal': False, 'reflection': False, 'title': 'Tools', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/websearchtool-error/160510/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229242, 'name': 'Damian Taubaso', 'username': 'dtaubaso', 'avatar_template': '/user_avatar/discuss.huggingface.co/dtaubaso/{size}/50040_2.png', 'created_at': '2025-06-24T20:34:07.645Z', 'cooked': 'I’m having a similar error with DuckDuckGo
\nCode execution failed at line ‘results_retry = web_search(query=simpler_query)’
\ndue to: DuckDuckGoSearchException: DuckDuckGo
\nRuntimeError: error sending request for url (DuckDuckGo):
\noperation timed out
Caused by:
\noperation timed out
Hmm… Perhaps DDG problem…?
\n\n\nOr perhaps:
\npip install -U duckduckgo-search\nHi, thanks for answering, I tried the StackOverflow solution already, the issue seems to be solved, but now I got max retries exceeded error, I still try to find the solution for it
', 'post_number': 5, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-06-26T10:51:24.636Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 160510, 'topic_slug': 'websearchtool-error', 'display_username': 'doradoradorayaki', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97781, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/websearchtool-error/160510/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229524, 'name': 'doradoradorayaki', 'username': 'dorayaki78', 'avatar_template': '/user_avatar/discuss.huggingface.co/dorayaki78/{size}/50008_2.png', 'created_at': '2025-06-26T10:52:55.396Z', 'cooked': 'have you figured out the solution yet, cause I solved the SSL issue already but stuck with the same problem as you
', 'post_number': 6, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-06-26T10:52:55.396Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 160510, 'topic_slug': 'websearchtool-error', 'display_username': 'doradoradorayaki', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 97828, 'username': 'dtaubaso', 'name': 'Damian Taubaso', 'avatar_template': '/user_avatar/discuss.huggingface.co/dtaubaso/{size}/50040_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97781, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/websearchtool-error/160510/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229533, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-26T12:41:36.577Z', 'cooked': 'Hmm… For example, how about with WebSearchTool(engine=""bing"") ?
I tried it, it is working now haha, at least it can surf the internet, but the result still need to be finetuned i think, thanks for the recommendation
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 9, 'post_type': 3, 'posts_count': 9, 'updated_at': '2025-06-29T01:06:38.554Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 160510, 'topic_slug': 'websearchtool-error', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/websearchtool-error/160510/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi I tried to use WebSearchTool from smolagents and got this kind of error, I’m using ollama with model qwen2.5 7b, can anyone help me
+Code execution failed at line ‘music_recommendations = web_search(query=“best party music”)’ due to: SSLError:
+HTTPSConnectionPool(host=‘lite.duckduckgo.com’, port=443): Max retries exceeded with url: /lite/?q=best+party+music
+(Caused by SSLError(SSLCertVerificationError(1, ‘[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed:
+self-signed certificate (_ssl.c:1028)’)))
Hmm… For example, how about with WebSearchTool(engine=""bing"") ?
Namely, I need a model that satisfies a few conditions, and one of them is that it has LlamaForCausalLM architecture. But I can’t find any interface that allows me to filter for such models, or list them. Any good ways to do this?
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-28T02:18:39.807Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 16, 'reads': 8, 'readers_count': 7, 'score': 91.6, 'yours': False, 'topic_id': 160965, 'topic_slug': 'how-can-i-search-models-by-architecture', 'display_username': 'Kim Byoungkwon', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98114, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-can-i-search-models-by-architecture/160965/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 229821, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-28T03:56:51.617Z', 'cooked': 'Since pipeline_tag is automatically assigned by Hugging Face Hub, it is possible to search by pipeline, but in the case of Transformers, pipeline_tag is determined by the task name, so there is currently no established method for searching by model architecture. Incidentally, in the case of Diffusers models, the architecture name is included in diffusers:, so it is possible.
If the model author has assigned tags themselves, you can search by specifying them with other=.
Searching with other=llama worked well enough for me, thank you so much!
There no direct filter for architecture. yet, search llama in the hub, it might work
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-28T12:09:39.891Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 16.6, 'yours': False, 'topic_id': 160965, 'topic_slug': 'how-can-i-search-models-by-architecture', 'display_username': 'Felicity Wood', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97008, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-can-i-search-models-by-architecture/160965/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229937, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-29T00:09:42.459Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-29T00:09:42.459Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 160965, 'topic_slug': 'how-can-i-search-models-by-architecture', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-can-i-search-models-by-architecture/160965/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Namely, I need a model that satisfies a few conditions, and one of them is that it has LlamaForCausalLM architecture. But I can’t find any interface that allows me to filter for such models, or list them. Any good ways to do this?
","Since pipeline_tag is automatically assigned by Hugging Face Hub, it is possible to search by pipeline, but in the case of Transformers, pipeline_tag is determined by the task name, so there is currently no established method for searching by model architecture. Incidentally, in the case of Diffusers models, the architecture name is included in diffusers:, so it is possible.
If the model author has assigned tags themselves, you can search by specifying them with other=.
Hello everyone,
\nI am trying to export the “Qwen/Qwen3-Embedding-0.6B” model to ONNX using the “optimum” library. According to the Optimum documentation, the “Qwen3” architecture is supported for ONNX export.
\nHowever, the export process fails with a error: “invalid unordered_map<K, T> key”
\nfrom optimum.exporters.onnx import main_export\nimport os\n\nmodel_id = ""Qwen/Qwen3-Embedding-0.6B""\noutput_dir = ""qwen3_embedding_onnx_from_script""\nos.makedirs(output_dir, exist_ok=True)\n\nprint(f""start export \'{model_id}\' "")\n\ntry:\n main_export(\n model_id,\n output=output_dir,\n task=""feature-extraction"",\n trust_remote_code=True,\n opset=20\n )\n print(f""Model \'{model_id}\' finish \'{output_dir}\'"")\n\nexcept Exception as e:\n print(f""error: {e}"")\ntask=\'feature-extraction\' and task=\'default\' (by letting optimum infer it automatically).invalid unordered_map<K, T> key error.This seems pretty difficult to get working. I failed too. I don’t want to reinstall PyTorch…
# pip install -U optimum[onnxruntime]\n# pip install -U accelerate transformers sentence-transformers\n\nfrom optimum.exporters.onnx import main_export\nimport os\n\nmodel_id = ""Qwen/Qwen3-Embedding-0.6B""\noutput_dir = ""qwen3_embedding_onnx_from_script""\nos.makedirs(output_dir, exist_ok=True)\n\nprint(f""start export \'{model_id}\' "")\n\ntry:\n main_export(\n model_id,\n output=output_dir,\n task=""feature-extraction"",\n trust_remote_code=True,\n opset=20 # opset=17 with PyTorch 1.x may work? https://huggingface.co/zhiqing/Qwen3-Embedding-0.6B-ONNX/discussions/1 https://github.com/pytorch/pytorch/issues/120559\n # With 2.x, ""error: Exporting the operator \'aten::__ior_\' to ONNX opset version 20 is not supported.""\n )\n print(f""Model \'{model_id}\' finish \'{output_dir}\'"")\n\nexcept Exception as e:\n print(f""error: {e}"")\n\n\n\n
invalid unordered_map<K, T> keyerror.
Seems 2.x issue, too…
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-27T15:00:01.857Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 8, 'readers_count': 7, 'score': 41.4, 'yours': False, 'topic_id': 160909, 'topic_slug': 'onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/onnx/onnx/issues/5862', 'internal': False, 'reflection': False, 'clicks': 6}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key/160909/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 229730, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-27T15:11:09.025Z', 'cooked': 'Probably, if a parameter that forces attn_implementation=""eager"" at model.from_pretrained() part is implemented in Exporter, it will work with PyTorch 2.x as well…
Thank you for your help! Unfortunately, your suggestions didn’t work:
\nIt seems the issue is deeper at the compatibility level between Qwen3 architecture and current PyTorch/ONNX versions. (((((
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-27T15:41:18.226Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 7, 'readers_count': 6, 'score': 21.2, 'yours': False, 'topic_id': 160909, 'topic_slug': 'onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key', 'display_username': 'Nikolskiy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98077, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key/160909/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229791, 'name': 'Nikolskiy', 'username': 'Colegero', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/c/eada6e/{size}.png', 'created_at': '2025-06-27T22:39:09.088Z', 'cooked': 'Yeah, the error was indeed tied to torch 2.6.0. I installed this combo: pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1, and the issue is gone—thanks for the heads-up! Man, I’m so fed up with these constant PyTorch “rollercoasters” (((
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-27T22:39:09.088Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 6, 'readers_count': 5, 'score': 36.0, 'yours': False, 'topic_id': 160909, 'topic_slug': 'onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key', 'display_username': 'Nikolskiy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 98077, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key/160909/5', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229861, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-28T10:40:04.437Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-06-28T10:40:04.437Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 16, 'reads': 5, 'readers_count': 4, 'score': 40.8, 'yours': False, 'topic_id': 160909, 'topic_slug': 'onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/onnx-export-failed-for-qwen-qwen3-embedding-0-6b-with-invalid-unordered-map-k-t-key/160909/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone,
+I am trying to export the “Qwen/Qwen3-Embedding-0.6B” model to ONNX using the “optimum” library. According to the Optimum documentation, the “Qwen3” architecture is supported for ONNX export.
+However, the export process fails with a error: “invalid unordered_map<K, T> key”
+from optimum.exporters.onnx import main_export
+import os
+
+model_id = ""Qwen/Qwen3-Embedding-0.6B""
+output_dir = ""qwen3_embedding_onnx_from_script""
+os.makedirs(output_dir, exist_ok=True)
+
+print(f""start export '{model_id}' "")
+
+try:
+ main_export(
+ model_id,
+ output=output_dir,
+ task=""feature-extraction"",
+ trust_remote_code=True,
+ opset=20
+ )
+ print(f""Model '{model_id}' finish '{output_dir}'"")
+
+except Exception as e:
+ print(f""error: {e}"")
+task='feature-extraction' and task='default' (by letting optimum infer it automatically).invalid unordered_map<K, T> key error.This seems pretty difficult to get working. I failed too. I don’t want to reinstall PyTorch…
# pip install -U optimum[onnxruntime]
+# pip install -U accelerate transformers sentence-transformers
+
+from optimum.exporters.onnx import main_export
+import os
+
+model_id = ""Qwen/Qwen3-Embedding-0.6B""
+output_dir = ""qwen3_embedding_onnx_from_script""
+os.makedirs(output_dir, exist_ok=True)
+
+print(f""start export '{model_id}' "")
+
+try:
+ main_export(
+ model_id,
+ output=output_dir,
+ task=""feature-extraction"",
+ trust_remote_code=True,
+ opset=20 # opset=17 with PyTorch 1.x may work? https://huggingface.co/zhiqing/Qwen3-Embedding-0.6B-ONNX/discussions/1 https://github.com/pytorch/pytorch/issues/120559
+ # With 2.x, ""error: Exporting the operator 'aten::__ior_' to ONNX opset version 20 is not supported.""
+ )
+ print(f""Model '{model_id}' finish '{output_dir}'"")
+
+except Exception as e:
+ print(f""error: {e}"")
++++
invalid unordered_map<K, T> keyerror.
Seems 2.x issue, too…
+" +Scheduling failure: unable to schedule,https://discuss.huggingface.co/t/scheduling-failure-unable-to-schedule/160642,160642,64,2025-06-25 14:19:57.042000+00:00,"[{'id': 229359, 'name': 'Alban Huntziger', 'username': 'Albaninho10', 'avatar_template': '/user_avatar/discuss.huggingface.co/albaninho10/{size}/50078_2.png', 'created_at': '2025-06-25T14:19:57.111Z', 'cooked': 'Hello,
\nI want to deploy my model but I always get this error after +/- 20 minutes of “deployment”:
\nEndpoint encountered an error.
\nYou can try restarting it using the “retry” button above. Check [ logs] for more details.
\n[Server message]Endpoint failed to start
\nScheduling failure: unable to schedule
And in the logs I get this error:
\nError 502 while fetching logs for ""mon-modele-bricks-hiv"":
Has this ever happened to anyone?
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-25T14:19:57.111Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 24, 'reads': 7, 'readers_count': 6, 'score': 181.4, 'yours': False, 'topic_id': 160642, 'topic_slug': 'scheduling-failure-unable-to-schedule', 'display_username': 'Alban Huntziger', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/ajay-hinduja-geneva-switzerland-swiss-scheduling-failure-unable-to-schedule-error/162031/2', 'internal': True, 'reflection': True, 'title': 'Ajay Hinduja Geneva, Switzerland (Swiss): ""Scheduling Failure: Unable to Schedule"" Error', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97887, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/scheduling-failure-unable-to-schedule/160642/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 229368, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-06-25T15:03:38.762Z', 'cooked': 'Hi @Albaninho10 Thank you for reporting! We’re investigating now.
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-25T15:03:38.762Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 160642, 'topic_slug': 'scheduling-failure-unable-to-schedule', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/scheduling-failure-unable-to-schedule/160642/2', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229578, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-06-26T20:18:28.866Z', 'cooked': 'Hi @Albaninho10 Thank you for waiting! This error message is related to availability of the GPU instance at the time of deployment - this can be resolved by selecting a different instance or region if possible.
\nWe’ve added updating this error message so that it’s clearer on the roadmap, though there’s no ETA just yet. Please let us know if you have any feedback about Inference Endpoints - we’re all ears!
\nI also wanted to mention our Model Catalog, which has ready-to-deploy models that require no additional customization and deployment is verified by Hugging Face.
\nLet us know if you have other questions.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-26T20:18:28.866Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 26.2, 'yours': False, 'topic_id': 160642, 'topic_slug': 'scheduling-failure-unable-to-schedule', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://endpoints.huggingface.co/catalog', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/scheduling-failure-unable-to-schedule/160642/3', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 229600, 'name': 'Andrew Scott', 'username': 'Pimpcat-AU', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png', 'created_at': '2025-06-27T00:31:07.836Z', 'cooked': 'I’ve seen similar issues with deployment failures related to GPU availability. From what you’re describing, it seems like the GPU instance may not be available when the model tries to deploy, which causes the 502 error. One possible solution is to try selecting a different instance type or region during deployment to ensure that there are available GPU resources at the time of deployment. Also, double check if there’s any region specific resource limitation that might be causing the issue.
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-27T00:31:33.137Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 160642, 'topic_slug': 'scheduling-failure-unable-to-schedule', 'display_username': 'Andrew Scott', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96276, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/scheduling-failure-unable-to-schedule/160642/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229660, 'name': 'Alban Huntziger', 'username': 'Albaninho10', 'avatar_template': '/user_avatar/discuss.huggingface.co/albaninho10/{size}/50078_2.png', 'created_at': '2025-06-27T07:44:09.723Z', 'cooked': 'Thanks for the reply, indeed by changing region and GPU the model is deployed correctly !
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-27T07:44:09.723Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 36.0, 'yours': False, 'topic_id': 160642, 'topic_slug': 'scheduling-failure-unable-to-schedule', 'display_username': 'Alban Huntziger', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 31941, 'username': 'meganariley', 'name': 'Megan Riley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97887, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/scheduling-failure-unable-to-schedule/160642/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229779, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-27T19:44:53.671Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-06-27T19:44:53.671Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 160642, 'topic_slug': 'scheduling-failure-unable-to-schedule', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/scheduling-failure-unable-to-schedule/160642/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+I want to deploy my model but I always get this error after +/- 20 minutes of “deployment”:
+Endpoint encountered an error.
+You can try restarting it using the “retry” button above. Check [ logs] for more details.
+[Server message]Endpoint failed to start
+Scheduling failure: unable to schedule
And in the logs I get this error:
+Error 502 while fetching logs for ""mon-modele-bricks-hiv"":
Has this ever happened to anyone?
","Hi @Albaninho10 Thank you for waiting! This error message is related to availability of the GPU instance at the time of deployment - this can be resolved by selecting a different instance or region if possible.
+We’ve added updating this error message so that it’s clearer on the roadmap, though there’s no ETA just yet. Please let us know if you have any feedback about Inference Endpoints - we’re all ears!
+I also wanted to mention our Model Catalog, which has ready-to-deploy models that require no additional customization and deployment is verified by Hugging Face.
+Let us know if you have other questions.
" +Inference result not aligned with local version of same model and revision,https://discuss.huggingface.co/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514,160514,64,2025-06-24 10:46:33.697000+00:00,"[{'id': 229141, 'name': 'Renaud Pelissier', 'username': 'rpelissier', 'avatar_template': '/user_avatar/discuss.huggingface.co/rpelissier/{size}/50013_2.png', 'created_at': '2025-06-24T10:46:33.757Z', 'cooked': 'Hello,
\nI am trying to run this embedding model “sentence-transformers/LaBSE” with revision=“836121a0533e5664b21c7aacc5d22951f2b8b25b” on the Inference Endpoints.
I have a result, but the embeddings numbers are different from the local execution. And not even correlated using cosine similarity.
\nAny idea what’s going on ?
\n\nfrom abc import ABC, abstractmethod\nimport numpy as np\nimport requests\nfrom sentence_transformers import SentenceTransformer\nfrom sbw_fiabilis.logger import get_logger, set_level\nimport os\nfrom dotenv import load_dotenv\n\nlogger = get_logger()\n\n\nclass EmbeddingInterface(ABC):\n """"""Interface abstraite pour les services d\'embedding.""""""\n \n @abstractmethod\n def encode(self, texts, batch_size=None, show_progress_bar=False):\n pass\n\n\nclass LocalEmbeddingService(EmbeddingInterface):\n """"""Implémentation locale utilisant SentenceTransformer.""""""\n \n def __init__(self):\n WORKING_DIR = os.getenv(""WORKING_DIR"", os.path.join(os.path.dirname(__file__), ""../../data/working_dir""))\n HF_HOME = os.path.join(WORKING_DIR, "".hf"")\n os.environ[""HF_HOME""] = HF_HOME\n\n self.model = SentenceTransformer(""sentence-transformers/LaBSE"", revision=""836121a0533e5664b21c7aacc5d22951f2b8b25b"", cache_folder=HF_HOME)\n logger.info(f""LocalEmbeddingService configuré"")\n \n def encode(self, texts, batch_size=32, show_progress_bar=False):\n return self.model.encode(texts, batch_size=batch_size, show_progress_bar=show_progress_bar)\n\n\nclass APIEmbeddingService(EmbeddingInterface):\n """"""Implémentation utilisant l\'API Hugging Face.""""""\n \n def __init__(self):\n self.api_url = os.getenv(""EMBEDDING_API_URL"")\n self.api_key = os.getenv(""EMBEDDING_API_KEY"")\n if not self.api_url or not self.api_key:\n raise ValueError(""EMBEDDING_API_URL et EMBEDDING_API_KEY doivent être définis"")\n self.headers = {\n ""Accept"": ""application/json"",\n ""Authorization"": f""Bearer {self.api_key}"",\n ""Content-Type"": ""application/json""\n }\n logger.info(f""ApiEmbeddingService configuré"")\n \n def _query_api(self, payload):\n try:\n response = requests.post(self.api_url, headers=self.headers, json=payload, timeout=30)\n response.raise_for_status()\n return response.json()\n except requests.exceptions.RequestException as e:\n logger.error(f""Erreur lors de la requête API: {e}"")\n raise\n \n def encode(self, texts, batch_size=32, show_progress_bar=False):\n if not texts:\n return np.array([])\n \n all_embeddings = []\n total_texts = len(texts)\n \n logger.info(f""Encodage via API: {total_texts} textes en lots de {batch_size}"")\n \n for i in range(0, total_texts, batch_size):\n batch = texts[i:i + batch_size]\n \n payload = {\n ""inputs"": batch,\n ""parameters"": {}\n }\n \n response = self._query_api(payload)\n \n # Gestion des différents formats de réponse API\n if isinstance(response, list):\n batch_embeddings = response\n elif isinstance(response, dict) and ""embeddings"" in response:\n batch_embeddings = response[""embeddings""]\n else:\n raise ValueError(f""Format de réponse API inattendu: {type(response)}"")\n \n all_embeddings.extend(batch_embeddings)\n \n logger.info(f"" Lot traité: {min(i + batch_size, total_texts)}/{total_texts}"")\n \n return all_embeddings\n\n\n\n\n\ndef test():\n logger = get_logger()\n set_level(""DEBUG"")\n\n load_dotenv()\n\n texts = [""toto"", ""tata""]\n\n service = LocalEmbeddingService()\n embeddings = service.encode(texts)\n logger.info(embeddings[0][:5])\n logger.info(embeddings[1][:5])\n\n service = APIEmbeddingService()\n embeddings = service.encode(texts)\n logger.info(embeddings[0][:5])\n logger.info(embeddings[1][:5])\n\nif __name__ == ""__main__"":\n test()\nThe result with different embeddings.
\nINFO - Logger level set to INFO\nINFO - Logger level set to DEBUG\nINFO - LocalEmbeddingService configuré\nINFO - [ 0.02300638 -0.07002795 -0.01850945 -0.03634194 0.0507826 ]\nINFO - [-0.03088209 -0.05037568 -0.00730146 -0.0068823 0.03126564]\nINFO - ApiEmbeddingService configuré\nINFO - Encodage via API: 2 textes en lots de 32\nINFO - Lot traité: 2/2\nINFO - [0.0077932924, 0.015989138, 0.010355308, 0.0026318827, 0.019499298]\nINFO - [-0.007399403, -0.03194063, -0.016836794, 0.022840464, 0.001694431]\nIf you select anything other than “Custom,” I think the contents of handler.py will be ignored. In this case, I think model will be executed with the default arguments of the default pipeline. That may be why there is a difference from the local code.
Thank you John for helping.
\nI am not using this way of running an endpoint, I am using the no-code approach and the UI is showing the right model with the right version (screenshots).
This means that either the library (in this case, TGI and SentenceTransformers) is installed locally or on the endpoint, or the code for the template is simply buggy…
\nIf the repository version specification does not work, that may also be a bug, but if that is the only issue, the cosine similarity should not be extremely off.
As shown below, a fairly old version of the library is used in the endpoint. Of course, it is possible to update it manually…
\n', 'post_number': 6, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-06-24T14:22:07.337Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 6.4, 'yours': False, 'topic_id': 160514, 'topic_slug': 'inference-result-not-aligned-with-local-version-of-same-model-and-revision', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/inference-endpoints/others/runtime', 'internal': False, 'reflection': False, 'title': 'Inference Endpoints Version', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229187, 'name': 'Renaud Pelissier', 'username': 'rpelissier', 'avatar_template': '/user_avatar/discuss.huggingface.co/rpelissier/{size}/50013_2.png', 'created_at': '2025-06-24T14:25:36.828Z', 'cooked': 'Indeed the log of the replica doesn’t really seems to take into account any of the params provided in the UI.
\nThe log of the replica :
\n\n', 'post_number': 7, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-06-24T14:26:16.484Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 160514, 'topic_slug': 'inference-result-not-aligned-with-local-version-of-same-model-and-revision', 'display_username': 'Renaud Pelissier', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97785, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229189, 'name': 'Renaud Pelissier', 'username': 'rpelissier', 'avatar_template': '/user_avatar/discuss.huggingface.co/rpelissier/{size}/50013_2.png', 'created_at': '2025-06-24T14:31:31.849Z', 'cooked': 'Args { model_id: “/rep****ory”, revision: None, tokenization_workers: None, dtype: None, pooling: None, max_concurrent_requests: 512, max_batch_tokens: 16384, max_batch_requests: None, max_client_batch_size: 32, auto_truncate: false, default_prompt_name: None, default_prompt: None, hf_api_token: None, hf_token: None, hostname: “r-rpelissier-sbw-fidi-labse-58w96y74-e4770-0t00y”, port: 80, uds_path: “/tmp/text-embeddings-inference-server”, huggingface_hub_cache: Some(“/repository/cache”), payload_limit: 2000000, api_key: None, json_output: true, disable_spans: false, otlp_endpoint: None, otlp_service_name: “text-embeddings-inference.server”, cors_allow_origin: None }
\n
Too bad, if I need to debug this (a paid service).
\nThe purpose of a managed service is to ignore the underlying complexity of provisioning, maintaining versions… I am really disappointed by what seems to be a “tools for POC” but not a production ready service.
\nAnd having a mailto:… (that attempt to open my mail desktop app instead of gmail) as the only way to reach the support was another proof that this is not too serious.
If it’s for a paid service, using Expert Support is probably the fastest and most reliable option, especially if it seems like a bug.
\n\nBTW, on my local PC:
\nfrom sentence_transformers import SentenceTransformer # sentence-transformers 4.0.1\nimport torch\nsentences = [""This is an example sentence"", ""Each sentence is converted""]\ndevice = ""cuda"" if torch.cuda.is_available() else ""cpu""\nprint(f""Running on {device}."") # Running on cuda.\n\nmodel = SentenceTransformer(""sentence-transformers/LaBSE"").to(device)\nembeddings = model.encode(sentences)\nprint(""main:"", embeddings)\n#main: [[ 0.02882478 -0.00602382 -0.05947006 ... -0.03002249 -0.029607\n# 0.00067482]\n# [-0.05550233 0.02546483 -0.02157256 ... 0.02932105 0.01150041\n# -0.00848792]]\n\nmodel = SentenceTransformer(""sentence-transformers/LaBSE"", revision=""836121a0533e5664b21c7aacc5d22951f2b8b25b"").to(device)\nembeddings = model.encode(sentences)\nprint(""836121a0533e5664b21c7aacc5d22951f2b8b25b:"", embeddings)\n#836121a0533e5664b21c7aacc5d22951f2b8b25b: [[ 0.02882478 -0.00602382 -0.05947006 ... -0.03002249 -0.029607\n# 0.00067482]\n# [-0.05550233 0.02546483 -0.02157256 ... 0.02932105 0.01150041\n# -0.00848792]]\n\nmodel.to(""cpu"")\nembeddings = model.encode(sentences)\nprint(""On CPU:"", embeddings)\n#On CPU: [[ 0.02882476 -0.00602385 -0.05947007 ... -0.03002251 -0.02960699\n# 0.00067482]\n# [-0.05550234 0.02546484 -0.02157255 ... 0.02932107 0.01150037\n# -0.00848786]]\nAt least locally consistent. Thank you !
', 'post_number': 10, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-06-24T15:03:39.346Z', 'reply_count': 0, 'reply_to_post_number': 9, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 160514, 'topic_slug': 'inference-result-not-aligned-with-local-version-of-same-model-and-revision', 'display_username': 'Renaud Pelissier', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97785, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514/10', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229349, 'name': 'Erik Kaunismäki', 'username': 'erikkaum', 'avatar_template': '/user_avatar/discuss.huggingface.co/erikkaum/{size}/29571_2.png', 'created_at': '2025-06-25T13:34:16.110Z', 'cooked': 'Hi rpelissier
Sorry about the hassle here. I did a deep dive on issue and I think I know what’s going on: the model deployed in your inference endpoint uses the TEI server engine. Whereas the local example uses sentence-transformers. And unfortunately there’s a mismatch between the implementations. This model is one of the few that uses a Dense module, which is supported in sentence transformers but not in TEI.
\nSo when the model is ran with TEI (and therefore on inference endpoints), it’s equivalent to doing this in sentence transformers:
\nfrom sentence_transformers import SentenceTransformer\nimport torch\nsentences = [""This is an example sentence"", ""Each sentence is converted""]\ndevice = ""cuda"" if torch.cuda.is_available() else ""cpu""\nprint(f""Running on {device}."")\n\nmodel = SentenceTransformer(""sentence-transformers/LaBSE"").to(device)\nembeddings = model.encode(sentences)\nprint(""default"", embeddings)\n\nedited_model = SentenceTransformer(""sentence-transformers/LaBSE"").to(device)\ndel edited_model[2]\nembeddings = edited_model.encode(sentences)\nprint(""del model[2]:"", embeddings)\nthis gives the output:
\ndefault [[ 0.02882483 -0.00602379 -0.05947006 ... -0.03002251 -0.029607\n 0.00067482]\n [-0.05550232 0.02546485 -0.02157257 ... 0.02932104 0.0115004\n -0.00848789]]\ndel model[2]: [[-0.00814162 0.01150823 -0.01516913 ... -0.02249936 0.02313923\n -0.02578063]\n [ 0.00584357 0.03796612 0.0039336 ... 0.03305857 0.03542801\n 0.0157448 ]]\nwhere the former corresponds to the same results in the post above, and the latter should be similar to the model deployed on inference endpoints with TEI.
\nThis is indeed not ideal and I’ve notified the maintainers of TEI so they can work on either supporting the Dense feature or alternatively clearly showing that this model isn���t supported in TEI.
\nAs a potential solution, when you deploy this model on Inference Endpoints, you can select the “Default” container instead of the TEI one. The default container is a simple wrapper around the sentence transformers library, so it’s not as performant as TEI, but it should give you the correct embeddings.
\n\nHopefully this helps
Thank you, erikkaum!
', 'post_number': 12, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-06-25T13:59:29.994Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 51.2, 'yours': False, 'topic_id': 160514, 'topic_slug': 'inference-result-not-aligned-with-local-version-of-same-model-and-revision', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514/12', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229506, 'name': 'Renaud Pelissier', 'username': 'rpelissier', 'avatar_template': '/user_avatar/discuss.huggingface.co/rpelissier/{size}/50013_2.png', 'created_at': '2025-06-26T09:08:21.026Z', 'cooked': 'Thank tou erikkaum, now I understand.
\nSo this feels like a serious bug to have an inference service ignoring some layers of the inference model. A big warning should show, at least.
\nI am sorry but to me it is a blocker for adoption of your product. It is a nice idea, but not reliable for production. I will give another try in 6 months. In the mean time I will go terraform and some autoscalable docker container. (No so easy though, I have been working on it for the past couple of day, and autoscaling with caching the model weights and with enough CPU, is not really what it was designed for.
Hi rpelissier,
\nI totally understand and agree that it’s a serious bug.
\nAlso just as a heads up: if you deploy this model on your own infra with the text-embedding-inference server, you’ll have the same bug.
\nSo when you deploy on your own infra make sure to use the sentence-transformer implementation so that the embeddings are correct
Hey @rpelissier thanks for reporting! We’ve just pushed the changes to fix that and handle the 2_Dense/ modules when available on the Hub, it’s still a work in progress at Add `Dense`, `DenseLayer` and `DenseConfig` to handle `2_Dense/` by alvarobartt · Pull Request #660 · huggingface/text-embeddings-inference · GitHub but we hope to release it soon, so stay tuned and we’ll ping you back
Also thanks a lot @erikkaum for handling, @tomaarsen for the assistance while solving it and @Narsil for the PR review!
', 'post_number': 15, 'post_type': 1, 'posts_count': 16, 'updated_at': '2025-06-26T16:33:19.049Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 76.2, 'yours': False, 'topic_id': 160514, 'topic_slug': 'inference-result-not-aligned-with-local-version-of-same-model-and-revision', 'display_username': 'Alvaro Bartolome', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/text-embeddings-inference/pull/660', 'internal': False, 'reflection': False, 'title': 'Add `Dense`, `DenseLayer` and `DenseConfig` to handle `2_Dense/` by alvarobartt · Pull Request #660 · huggingface/text-embeddings-inference · GitHub', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 4853, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514/15', 'reactions': [{'id': 'clap', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229668, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-27T08:24:30.058Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 16, 'post_type': 3, 'posts_count': 16, 'updated_at': '2025-06-27T08:24:30.058Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 160514, 'topic_slug': 'inference-result-not-aligned-with-local-version-of-same-model-and-revision', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/inference-result-not-aligned-with-local-version-of-same-model-and-revision/160514/16', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+I am trying to run this embedding model “sentence-transformers/LaBSE” with revision=“836121a0533e5664b21c7aacc5d22951f2b8b25b” on the Inference Endpoints.
I have a result, but the embeddings numbers are different from the local execution. And not even correlated using cosine similarity.
+Any idea what’s going on ?
+ +from abc import ABC, abstractmethod
+import numpy as np
+import requests
+from sentence_transformers import SentenceTransformer
+from sbw_fiabilis.logger import get_logger, set_level
+import os
+from dotenv import load_dotenv
+
+logger = get_logger()
+
+
+class EmbeddingInterface(ABC):
+ """"""Interface abstraite pour les services d'embedding.""""""
+
+ @abstractmethod
+ def encode(self, texts, batch_size=None, show_progress_bar=False):
+ pass
+
+
+class LocalEmbeddingService(EmbeddingInterface):
+ """"""Implémentation locale utilisant SentenceTransformer.""""""
+
+ def __init__(self):
+ WORKING_DIR = os.getenv(""WORKING_DIR"", os.path.join(os.path.dirname(__file__), ""../../data/working_dir""))
+ HF_HOME = os.path.join(WORKING_DIR, "".hf"")
+ os.environ[""HF_HOME""] = HF_HOME
+
+ self.model = SentenceTransformer(""sentence-transformers/LaBSE"", revision=""836121a0533e5664b21c7aacc5d22951f2b8b25b"", cache_folder=HF_HOME)
+ logger.info(f""LocalEmbeddingService configuré"")
+
+ def encode(self, texts, batch_size=32, show_progress_bar=False):
+ return self.model.encode(texts, batch_size=batch_size, show_progress_bar=show_progress_bar)
+
+
+class APIEmbeddingService(EmbeddingInterface):
+ """"""Implémentation utilisant l'API Hugging Face.""""""
+
+ def __init__(self):
+ self.api_url = os.getenv(""EMBEDDING_API_URL"")
+ self.api_key = os.getenv(""EMBEDDING_API_KEY"")
+ if not self.api_url or not self.api_key:
+ raise ValueError(""EMBEDDING_API_URL et EMBEDDING_API_KEY doivent être définis"")
+ self.headers = {
+ ""Accept"": ""application/json"",
+ ""Authorization"": f""Bearer {self.api_key}"",
+ ""Content-Type"": ""application/json""
+ }
+ logger.info(f""ApiEmbeddingService configuré"")
+
+ def _query_api(self, payload):
+ try:
+ response = requests.post(self.api_url, headers=self.headers, json=payload, timeout=30)
+ response.raise_for_status()
+ return response.json()
+ except requests.exceptions.RequestException as e:
+ logger.error(f""Erreur lors de la requête API: {e}"")
+ raise
+
+ def encode(self, texts, batch_size=32, show_progress_bar=False):
+ if not texts:
+ return np.array([])
+
+ all_embeddings = []
+ total_texts = len(texts)
+
+ logger.info(f""Encodage via API: {total_texts} textes en lots de {batch_size}"")
+
+ for i in range(0, total_texts, batch_size):
+ batch = texts[i:i + batch_size]
+
+ payload = {
+ ""inputs"": batch,
+ ""parameters"": {}
+ }
+
+ response = self._query_api(payload)
+
+ # Gestion des différents formats de réponse API
+ if isinstance(response, list):
+ batch_embeddings = response
+ elif isinstance(response, dict) and ""embeddings"" in response:
+ batch_embeddings = response[""embeddings""]
+ else:
+ raise ValueError(f""Format de réponse API inattendu: {type(response)}"")
+
+ all_embeddings.extend(batch_embeddings)
+
+ logger.info(f"" Lot traité: {min(i + batch_size, total_texts)}/{total_texts}"")
+
+ return all_embeddings
+
+
+
+
+
+def test():
+ logger = get_logger()
+ set_level(""DEBUG"")
+
+ load_dotenv()
+
+ texts = [""toto"", ""tata""]
+
+ service = LocalEmbeddingService()
+ embeddings = service.encode(texts)
+ logger.info(embeddings[0][:5])
+ logger.info(embeddings[1][:5])
+
+ service = APIEmbeddingService()
+ embeddings = service.encode(texts)
+ logger.info(embeddings[0][:5])
+ logger.info(embeddings[1][:5])
+
+if __name__ == ""__main__"":
+ test()
+Hi rpelissier
Sorry about the hassle here. I did a deep dive on issue and I think I know what’s going on: the model deployed in your inference endpoint uses the TEI server engine. Whereas the local example uses sentence-transformers. And unfortunately there’s a mismatch between the implementations. This model is one of the few that uses a Dense module, which is supported in sentence transformers but not in TEI.
+So when the model is ran with TEI (and therefore on inference endpoints), it’s equivalent to doing this in sentence transformers:
+from sentence_transformers import SentenceTransformer
+import torch
+sentences = [""This is an example sentence"", ""Each sentence is converted""]
+device = ""cuda"" if torch.cuda.is_available() else ""cpu""
+print(f""Running on {device}."")
+
+model = SentenceTransformer(""sentence-transformers/LaBSE"").to(device)
+embeddings = model.encode(sentences)
+print(""default"", embeddings)
+
+edited_model = SentenceTransformer(""sentence-transformers/LaBSE"").to(device)
+del edited_model[2]
+embeddings = edited_model.encode(sentences)
+print(""del model[2]:"", embeddings)
+this gives the output:
+default [[ 0.02882483 -0.00602379 -0.05947006 ... -0.03002251 -0.029607
+ 0.00067482]
+ [-0.05550232 0.02546485 -0.02157257 ... 0.02932104 0.0115004
+ -0.00848789]]
+del model[2]: [[-0.00814162 0.01150823 -0.01516913 ... -0.02249936 0.02313923
+ -0.02578063]
+ [ 0.00584357 0.03796612 0.0039336 ... 0.03305857 0.03542801
+ 0.0157448 ]]
+where the former corresponds to the same results in the post above, and the latter should be similar to the model deployed on inference endpoints with TEI.
+This is indeed not ideal and I’ve notified the maintainers of TEI so they can work on either supporting the Dense feature or alternatively clearly showing that this model isn’t supported in TEI.
+As a potential solution, when you deploy this model on Inference Endpoints, you can select the “Default” container instead of the TEI one. The default container is a simple wrapper around the sentence transformers library, so it’s not as performant as TEI, but it should give you the correct embeddings.
+ +Hopefully this helps
Hey, do you know current models that can also be executed locally, i.e. not in the cloud
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-24T13:24:11.780Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 242, 'reads': 10, 'readers_count': 9, 'score': 1157.0, 'yours': False, 'topic_id': 160530, 'topic_slug': 'what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time', 'display_username': 'Dizzy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97797, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time/160530/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 229166, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-24T13:33:46.771Z', 'cooked': 'When it comes to locally executable models, the Whisper series seems to have a lot of know-how. However, there are other options as well.
\nIn terms of speed, FastRTC excels in real-time performance, but it’s quite specialized. Or rather, it’s cloud-based?
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-24T13:34:00.248Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 10, 'readers_count': 9, 'score': 62.0, 'yours': False, 'topic_id': 160530, 'topic_slug': 'what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/hf-audio/open_asr_leaderboard', 'internal': False, 'reflection': False, 'title': 'Open ASR Leaderboard - a Hugging Face Space by hf-audio', 'clicks': 50}, {'url': 'https://github.com/gradio-app/fastrtc', 'internal': False, 'reflection': False, 'title': 'GitHub - gradio-app/fastrtc: The python library for real-time communication', 'clicks': 8}, {'url': 'https://huggingface.co/spaces?sort=trending&search=asr', 'internal': False, 'reflection': False, 'title': 'Spaces - Hugging Face', 'clicks': 5}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time/160530/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 229304, 'name': 'Dizzy', 'username': 'Dizzy22', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/9fc29f/{size}.png', 'created_at': '2025-06-25T06:49:23.774Z', 'cooked': 'Yes, I already have Whisper on my shortlist and it seems to be the best option. I’ve also heard about
\nDo you have any experience with these?
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-25T06:51:10.213Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 160530, 'topic_slug': 'what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time', 'display_username': 'Dizzy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97797, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time/160530/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229326, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-25T10:24:00.941Z', 'cooked': '\n\nDo you have any experience with these?
\n
No.
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-25T10:24:00.941Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 26.0, 'yours': False, 'topic_id': 160530, 'topic_slug': 'what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time/160530/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 229479, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-26T07:20:22.681Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-26T07:20:22.681Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 160530, 'topic_slug': 'what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/what-are-the-latest-open-source-speech-to-text-models-with-a-focus-on-real-time/160530/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hey, do you know current models that can also be executed locally, i.e. not in the cloud
","When it comes to locally executable models, the Whisper series seems to have a lot of know-how. However, there are other options as well.
+In terms of speed, FastRTC excels in real-time performance, but it’s quite specialized. Or rather, it’s cloud-based?
+ + +" +Unauthorized Access Token,https://discuss.huggingface.co/t/unauthorized-access-token/160609,160609,5,2025-06-25 09:01:15.843000+00:00,"[{'id': 229317, 'name': 'Philip Mockridge', 'username': 'FreeRoss', 'avatar_template': '/user_avatar/discuss.huggingface.co/freeross/{size}/50057_2.png', 'created_at': '2025-06-25T09:01:15.929Z', 'cooked': 'Hi,
\nThanks in advance if you’re able to help out.
\ncurl -H ""Authorization: Bearer hf_<...>bfQ"" https://huggingface.co/api/whoami\n{""error"":""Invalid credentials in Authorization header""}\nProvide the version of the library you are using:
\nI’m not using a library for this
If you have tried something in particular to solve your problem, don’t hesitate to mention it as well:
\nI tried to use the credentials initially in an n8n workflow → http request node. The curl is the simplest way to express this problem.
\nPlease find attached shot of the tokens I setup:
\n
The error message is clear as to what the problem is (unauthorized). What I do not know is why and/or why Huggingface server interprets the access token as anauthorized?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-25T09:01:15.929Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 38, 'reads': 11, 'readers_count': 10, 'score': 197.2, 'yours': False, 'topic_id': 160609, 'topic_slug': 'unauthorized-access-token', 'display_username': 'Philip Mockridge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97862, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unauthorized-access-token/160609/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 229325, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-25T10:22:46.004Z', 'cooked': 'Try v2.
\nHF_TOKEN = ""hf_foobar""\nimport subprocess\nsubprocess.run(f\'curl -H ""Authorization: Bearer {HF_TOKEN}"" https://huggingface.co/api/whoami\', shell=True)\n# {""error"":""Invalid credentials in Authorization header""}\nsubprocess.run(f\'curl -H ""Authorization: Bearer {HF_TOKEN}"" https://huggingface.co/api/whoami-v2\', shell=True)\n# {""type"":""user"", ...\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-06-26T05:47:53.399Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.4, 'yours': False, 'topic_id': 160609, 'topic_slug': 'unauthorized-access-token', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/unauthorized-access-token/160609/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+Thanks in advance if you’re able to help out.
+curl -H ""Authorization: Bearer hf_<...>bfQ"" https://huggingface.co/api/whoami
+{""error"":""Invalid credentials in Authorization header""}
+Provide the version of the library you are using:
+I’m not using a library for this
If you have tried something in particular to solve your problem, don’t hesitate to mention it as well:
+I tried to use the credentials initially in an n8n workflow → http request node. The curl is the simplest way to express this problem.
+Please find attached shot of the tokens I setup:
+
The error message is clear as to what the problem is (unauthorized). What I do not know is why and/or why Huggingface server interprets the access token as anauthorized?
","Try v2.
+HF_TOKEN = ""hf_foobar""
+import subprocess
+subprocess.run(f'curl -H ""Authorization: Bearer {HF_TOKEN}"" https://huggingface.co/api/whoami', shell=True)
+# {""error"":""Invalid credentials in Authorization header""}
+subprocess.run(f'curl -H ""Authorization: Bearer {HF_TOKEN}"" https://huggingface.co/api/whoami-v2', shell=True)
+# {""type"":""user"", ...
+The doc suggests that installing with the commands:
\npip install \'transformers[torch]\'\nuv pip install \'transformers[torch]\'\nwill get a CPU-only install (I don’t have a GPU). So why does it have to take >2GB of my disk space for CUDA-specific libraries? especially if I’m going to run this in a docker-type environment, I’m interested to know if it’s possible to install without the GBs of CUDA libraries. If that breaks the transformers functionality, I would be interested in editing the docs accordingly.
\nI do realize that it’s getting installed because of the torch, not because of transformers itself, but it would be nice to know if there’s a way to slim this down when it’s not needed.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-20T17:30:57.867Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 114, 'reads': 7, 'readers_count': 6, 'score': 556.4, 'yours': False, 'topic_id': 160110, 'topic_slug': 'why-does-installing-cpu-only-version-of-transformers-install-multiple-gb-of-cuda-libs', 'display_username': 'Faaiz Memon', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/en/installation?cpu-only=PyTorch#python', 'internal': False, 'reflection': False, 'title': 'Installation', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90281, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/why-does-installing-cpu-only-version-of-transformers-install-multiple-gb-of-cuda-libs/160110/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 228661, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-21T00:58:16.025Z', 'cooked': 'The Transoformers library also works with PyTorch for CPUs. However, if you install CUDA and then run pip install torch, the CUDA version will be installed. I think you can make it slimmer by installing PyTorch for CPU first somehow, and then installing Transoformers with pip install transoformers.
\nhttps://stackoverflow.com/questions/78947332/how-to-install-torch-without-nvidia
\nhttps://stackoverflow.com/questions/51730880/where-do-i-get-a-cpu-only-version-of-pytorch
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-06-24T14:31:22.261Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 160110, 'topic_slug': 'why-does-installing-cpu-only-version-of-transformers-install-multiple-gb-of-cuda-libs', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/why-does-installing-cpu-only-version-of-transformers-install-multiple-gb-of-cuda-libs/160110/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","The doc suggests that installing with the commands:
+pip install 'transformers[torch]'
+uv pip install 'transformers[torch]'
+will get a CPU-only install (I don’t have a GPU). So why does it have to take >2GB of my disk space for CUDA-specific libraries? especially if I’m going to run this in a docker-type environment, I’m interested to know if it’s possible to install without the GBs of CUDA libraries. If that breaks the transformers functionality, I would be interested in editing the docs accordingly.
+I do realize that it’s getting installed because of the torch, not because of transformers itself, but it would be nice to know if there’s a way to slim this down when it’s not needed.
","The Transoformers library also works with PyTorch for CPUs. However, if you install CUDA and then run pip install torch, the CUDA version will be installed. I think you can make it slimmer by installing PyTorch for CPU first somehow, and then installing Transoformers with pip install transoformers.
+https://stackoverflow.com/questions/78947332/how-to-install-torch-without-nvidia
+https://stackoverflow.com/questions/51730880/where-do-i-get-a-cpu-only-version-of-pytorch
Hi,
\nI’m new to HF dataset and I tried to create datasets based on data versioned in lakeFS (MinIO S3 bucket as storage backend)
\nHere I’m using ±30000 PIL image from MNIST data however it is taking around 12min to execute, which is a lot!
\nFrom what I understand, it is loading the images into cache then building the dataset.
\n– Please find bellow the execution screenshot –
Is there a way to optimize this or am I doing something wrong?
\n', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-19T11:58:46.893Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 31, 'reads': 8, 'readers_count': 7, 'score': 171.6, 'yours': False, 'topic_id': 159955, 'topic_slug': 'creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time', 'display_username': 'Adam BEN KHALIFA', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97330, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time/159955/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 228381, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-19T12:45:45.961Z', 'cooked': 'Hmm… There is not much information available.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-19T12:45:45.961Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 7, 'readers_count': 6, 'score': 21.4, 'yours': False, 'topic_id': 159955, 'topic_slug': 'creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/datasets/issues/6478', 'internal': False, 'reflection': False, 'title': 'How to load data from lakefs · Issue #6478 · huggingface/datasets · GitHub', 'clicks': 3}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time/159955/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228459, 'name': 'not-lain', 'username': 'not-lain', 'avatar_template': '/user_avatar/discuss.huggingface.co/not-lain/{size}/23122_2.png', 'created_at': '2025-06-19T22:53:55.820Z', 'cooked': '@Adam-Ben-Khalifa you can try to load the data in streaming mode, also after you converted the data into the datasets library consider saving it locally or pushing it to the hub
', 'post_number': 3, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-19T22:53:55.820Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 36.4, 'yours': False, 'topic_id': 159955, 'topic_slug': 'creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time', 'display_username': 'not-lain', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 38692, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time/159955/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228562, 'name': 'Adam BEN KHALIFA', 'username': 'Adam-Ben-Khalifa', 'avatar_template': '/user_avatar/discuss.huggingface.co/adam-ben-khalifa/{size}/49687_2.png', 'created_at': '2025-06-20T11:04:13.918Z', 'cooked': 'I’m saving the dataset locally, the delay is only at the first time creating it.
\nAlso I tried streaming and multiprocessing but I’m not seeing a difference, take a look
imagefolder is mainly for small image datasets, so I don’t think it’s very fast.
This is helpful, I didn’t see these posts since I didn’t consider the data I’m testing with large (around 30k images ~ 9MB total)
\nI’ll check them and post an update
\nThanks!
The bottleneck, from what I understand, was making one network request per file
\nFor 30k images, this meant 30k separate GET requests to the MinIO server through the S3 API, and that was killing the performance
\nUsing webDataset to transform the large number of files to few .tar files and passing “webdataset” instead of “imagefolder” to the load_dataset function worked perfectly (took only ~11s)
', 'post_number': 8, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-23T12:37:39.183Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 4, 'readers_count': 3, 'score': 40.8, 'yours': False, 'topic_id': 159955, 'topic_slug': 'creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time', 'display_username': 'Adam BEN KHALIFA', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97330, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time/159955/8', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 229046, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-24T00:37:45.162Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 9, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-06-24T00:37:45.162Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 159955, 'topic_slug': 'creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/creating-a-hf-dataset-from-lakefs-with-s3-storage-takes-too-much-time/159955/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I’m new to HF dataset and I tried to create datasets based on data versioned in lakeFS (MinIO S3 bucket as storage backend)
+Here I’m using ±30000 PIL image from MNIST data however it is taking around 12min to execute, which is a lot!
+From what I understand, it is loading the images into cache then building the dataset.
+– Please find bellow the execution screenshot –
Is there a way to optimize this or am I doing something wrong?
+",This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
+MCP Server Not Starting Despite GRADIO_MCP_SERVER=True in Gradio 5.27.1+,https://discuss.huggingface.co/t/mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1/160132,160132,21,2025-06-20 22:52:02.647000+00:00,"[{'id': 228653, 'name': 'usman fawad', 'username': 'usman69', 'avatar_template': '/user_avatar/discuss.huggingface.co/usman69/{size}/49822_2.png', 'created_at': '2025-06-20T22:52:02.733Z', 'cooked': 'I’m trying to expose my Gradio interface as an MCP server using the latest gradio[mcp] package (version 5.27.1). I’ve followed all the instructions in the MCP course and docs, including setting the environment variable before execution:
$env:GRADIO_MCP_SERVER=""True""\npy app.py\nHowever, the server only outputs:
\nRunning on local URL: http://127.0.0.1:7860\nand I never see the expected line:
\nMCP server available at: http://127.0.0.1:7860/gradio_api/mcp/sse\nI confirmed:
\ngradio==5.27.1 is installedgradio-mcp is also installedmcp_server=True in .launch() (since it’s removed in v5)py and python after setting the environment variableStill, the MCP server routes /gradio_api/mcp/sse and /schema never activate.
Could someone from the Gradio or MCP team help confirm if this is a bug or if something changed in v5 that isn’t reflected in the documentation?
\nReference: Building the Gradio MCP Server - Hugging Face MCP Course
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-20T22:53:23.192Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 158, 'reads': 12, 'readers_count': 11, 'score': 792.4, 'yours': False, 'topic_id': 160132, 'topic_slug': 'mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1', 'display_username': 'usman fawad', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/learn/mcp-course/unit2/gradio-server', 'internal': False, 'reflection': False, 'title': 'Building the Gradio MCP Server - Hugging Face MCP Course', 'clicks': 6}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97500, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1/160132/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 228668, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-21T01:34:23.344Z', 'cooked': 'Hmm… Perhaps this case?
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-21T01:34:23.344Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 10, 'reads': 10, 'readers_count': 9, 'score': 67.0, 'yours': False, 'topic_id': 160132, 'topic_slug': 'mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/gradio-app/gradio/issues/11225', 'internal': False, 'reflection': False, 'title': 'Erro while Connectin MCP server · Issue #11225 · gradio-app/gradio · GitHub', 'clicks': 11}, {'url': 'https://huggingface.co/spaces/abidlabs/mcp_tools2', 'internal': False, 'reflection': False, 'title': 'mcp_tools - a Hugging Face Space by abidlabs', 'clicks': 10}, {'url': 'https://github.com/gradio-app/gradio/issues/11225#issuecomment-2893381049', 'internal': False, 'reflection': False, 'title': 'Erro while Connectin MCP server · Issue #11225 · gradio-app/gradio · GitHub', 'clicks': 1}, {'url': 'https://github.com/abidlabs', 'internal': False, 'reflection': False, 'title': 'abidlabs (Abubakar Abid) · GitHub', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1/160132/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 228737, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-21T16:06:35.150Z', 'cooked': 'abidlabs
\n
\non May 20, 2025
\nOk I’ve figured out the issue, it’s due to a breaking change introduced by themcppackage going frommcp==1.8.1tomcp==1.9.0. We’re going to be investigating further to figure out if this breaking change inmcpis intentional or a mistake, but in the meantime, I recommend pinningmcp==1.8.1as in this Space: mcp_tools - a Hugging Face Space by abidlabs
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-06-21T16:06:35.150Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 6, 'readers_count': 5, 'score': 31.2, 'yours': False, 'topic_id': 160132, 'topic_slug': 'mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/mcp-server-not-starting-despite-gradio-mcp-server-true-in-gradio-5-27-1/160132/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m trying to expose my Gradio interface as an MCP server using the latest gradio[mcp] package (version 5.27.1). I’ve followed all the instructions in the MCP course and docs, including setting the environment variable before execution:
$env:GRADIO_MCP_SERVER=""True""
+py app.py
+However, the server only outputs:
+Running on local URL: http://127.0.0.1:7860
+and I never see the expected line:
+MCP server available at: http://127.0.0.1:7860/gradio_api/mcp/sse
+I confirmed:
+gradio==5.27.1 is installedgradio-mcp is also installedmcp_server=True in .launch() (since it’s removed in v5)py and python after setting the environment variableStill, the MCP server routes /gradio_api/mcp/sse and /schema never activate.
Could someone from the Gradio or MCP team help confirm if this is a bug or if something changed in v5 that isn’t reflected in the documentation?
+Reference: Building the Gradio MCP Server - Hugging Face MCP Course
","Hmm… Perhaps this case?
+ ++" +Make “image” column appear first in dataset preview UI,https://discuss.huggingface.co/t/make-image-column-appear-first-in-dataset-preview-ui/159787,159787,10,2025-06-18 09:22:03.753000+00:00,"[{'id': 228129, 'name': 'Cerveto Serrano', 'username': 'joancervetoserrano', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/82dd89/{size}.png', 'created_at': '2025-06-18T09:22:03.820Z', 'cooked': 'abidlabs
+
+on May 20, 2025
+Ok I’ve figured out the issue, it’s due to a breaking change introduced by themcppackage going frommcp==1.8.1tomcp==1.9.0. We’re going to be investigating further to figure out if this breaking change inmcpis intentional or a mistake, but in the meantime, I recommend pinningmcp==1.8.1as in this Space: mcp_tools - a Hugging Face Space by abidlabs
Hi!
I’m currently uploading a dataset that includes an ""image"" column (PNG files), along with some metadata columns. The dataset is loaded from a .jsonl file. My goal is to have the ""image"" column appear as the first column in the dataset card preview UI on the Hub.
However, at the moment, the ""image"" column is not the first—in fact, it appears last, which is not ideal for the presentation I’d like to achieve.
I have a couple of questions:
\n""image"" column first?.jsonl file or the features argument affect the display order?Thanks again for your time and help!
\n\nDoes the order of keys in the
\n.jsonlfile or thefeaturesargument affect the display order?
That’s probably true for datasets that have been loaded and saved in the datasets library.
However, if you simply upload image files as-is, I believe the order information will be automatically supplemented, so if you want to maintain the order in the viewer, you may need to manually create a settings file.
\nThe most reliable method is to convert the data to the parquet format using the datasets library (simply load and save).
Thank you!! I will check it!
\n
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-19T07:02:17.819Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 159787, 'topic_slug': 'make-image-column-appear-first-in-dataset-preview-ui', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/make-image-column-appear-first-in-dataset-preview-ui/159787/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi!
I’m currently uploading a dataset that includes an ""image"" column (PNG files), along with some metadata columns. The dataset is loaded from a .jsonl file. My goal is to have the ""image"" column appear as the first column in the dataset card preview UI on the Hub.
However, at the moment, the ""image"" column is not the first—in fact, it appears last, which is not ideal for the presentation I’d like to achieve.
I have a couple of questions:
+""image"" column first?.jsonl file or the features argument affect the display order?Thanks again for your time and help!
++Does the order of keys in the
+.jsonlfile or thefeaturesargument affect the display order?
That’s probably true for datasets that have been loaded and saved in the datasets library.
However, if you simply upload image files as-is, I believe the order information will be automatically supplemented, so if you want to maintain the order in the viewer, you may need to manually create a settings file.
+The most reliable method is to convert the data to the parquet format using the datasets library (simply load and save).
Seems like a silly question, but I’m learning and can’t find anything definitive…
\nIn models where input_ids and labels may be of different length (i.e. denoising, where a span of several tokens in labels may have been replaced by a single token), should the attention_mask correspond to labels (so the original chunk size) or to input_ids (so resized after noising)?
The attention_mask tells the model which positions in the input to attend to, i.e., which tokens are real vs padding. It applies only to the forward pass — specifically, how attention is computed over the input_ids.
\nThe labels are not used during attention computation — they are only used in the loss computation
', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-18T16:22:57.025Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 10, 'readers_count': 9, 'score': 37.0, 'yours': False, 'topic_id': 159820, 'topic_slug': 'does-attention-mask-refer-to-input-ids-or-to-labels', 'display_username': 'Riley Fox', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94214, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': 'Automatically removed quote of whole previous post.', 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/does-attention-mask-refer-to-input-ids-or-to-labels/159820/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228183, 'name': 'Philo Math', 'username': 'Philomath868', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/p/b487fb/{size}.png', 'created_at': '2025-06-18T16:41:13.944Z', 'cooked': 'Thanks, that’s a clear and succinct explanation!
\nBut I guess my question can still stand regarding decoder_input_ids, in case it’s based on labels (see my other question, which would mean - if I understand correctly - that labels (shifted right) are used during computation, at decoder side, no?
My bad, I completely didn’t see that
\nYes, the decoder_attention_mask (or just attention_mask on decoder_input_ids ) should match the decoder input, which is usually labels shifted right.
\ndecoder_input_ids are either provided manually or auto-generated by shifting labels right.
', 'post_number': 4, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-18T17:06:29.282Z', 'reply_count': 1, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 36.6, 'yours': False, 'topic_id': 159820, 'topic_slug': 'does-attention-mask-refer-to-input-ids-or-to-labels', 'display_username': 'Riley Fox', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 97307, 'username': 'Philomath868', 'name': 'Philo Math', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/p/b487fb/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94214, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/does-attention-mask-refer-to-input-ids-or-to-labels/159820/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228191, 'name': 'Philo Math', 'username': 'Philomath868', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/p/b487fb/{size}.png', 'created_at': '2025-06-18T17:13:17.484Z', 'cooked': 'So in my dataset, I should include both attention_mask and decoder_attention_mask? Will the model know which mask to use at which phase? I’m a bit confused…
', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-18T17:13:17.484Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 21.2, 'yours': False, 'topic_id': 159820, 'topic_slug': 'does-attention-mask-refer-to-input-ids-or-to-labels', 'display_username': 'Philo Math', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 94214, 'username': 'Mdrnfox', 'name': 'Riley Fox', 'avatar_template': '/user_avatar/discuss.huggingface.co/mdrnfox/{size}/47695_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97307, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/does-attention-mask-refer-to-input-ids-or-to-labels/159820/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228196, 'name': 'Riley Fox', 'username': 'Mdrnfox', 'avatar_template': '/user_avatar/discuss.huggingface.co/mdrnfox/{size}/47695_2.png', 'created_at': '2025-06-18T17:33:29.409Z', 'cooked': 'With HF Trainer, you only need to pass input_ids, attention_mask, labels
\nIf you pass labels, the model will:
\n1.\tAutomatically shift them to create decoder_input_ids
\n2.\tCreate the decoder_attention_mask to match the decoder_input_ids
\n3.\tHandle masking and loss computation (ignoring -100 in labels)
So the full decoder setup is inferred internally — as long as you provide labels.
\nYou do not need to manually include decoder_input_ids or decoder_attention_mask — they are automatically derived at runtime by the model or tokenizer.
', 'post_number': 6, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-06-18T17:33:29.575Z', 'reply_count': 1, 'reply_to_post_number': 5, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 6, 'readers_count': 5, 'score': 36.2, 'yours': False, 'topic_id': 159820, 'topic_slug': 'does-attention-mask-refer-to-input-ids-or-to-labels', 'display_username': 'Riley Fox', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 97307, 'username': 'Philomath868', 'name': 'Philo Math', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/p/b487fb/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94214, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': 'Automatically removed quote of whole previous post.', 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/does-attention-mask-refer-to-input-ids-or-to-labels/159820/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 228199, 'name': 'Philo Math', 'username': 'Philomath868', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/p/b487fb/{size}.png', 'created_at': '2025-06-18T17:40:16.713Z', 'cooked': 'Thank you!
\nSo just to make it absolutely clear (just correct me if I’m wrong; ignore otherwise ): I must pass attention_mask based on the noised text (input_ids), for the encoder. I can just leave the (possibly longer) decoder_attention_mask for the trainer to handle. Great!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-06-19T05:40:33.060Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 159820, 'topic_slug': 'does-attention-mask-refer-to-input-ids-or-to-labels', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/does-attention-mask-refer-to-input-ids-or-to-labels/159820/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Seems like a silly question, but I’m learning and can’t find anything definitive…
+In models where input_ids and labels may be of different length (i.e. denoising, where a span of several tokens in labels may have been replaced by a single token), should the attention_mask correspond to labels (so the original chunk size) or to input_ids (so resized after noising)?
With HF Trainer, you only need to pass input_ids, attention_mask, labels
+If you pass labels, the model will:
+1. Automatically shift them to create decoder_input_ids
+2. Create the decoder_attention_mask to match the decoder_input_ids
+3. Handle masking and loss computation (ignoring -100 in labels)
So the full decoder setup is inferred internally — as long as you provide labels.
+You do not need to manually include decoder_input_ids or decoder_attention_mask — they are automatically derived at runtime by the model or tokenizer.
" +Not seeing memory benefit to accelerate/FSDP2,https://discuss.huggingface.co/t/not-seeing-memory-benefit-to-accelerate-fsdp2/158039,158039,18,2025-06-04 21:34:41.903000+00:00,"[{'id': 225715, 'name': 'hpcpony', 'username': 'hpcpony', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/h/779978/{size}.png', 'created_at': '2025-06-04T21:34:41.982Z', 'cooked': 'TL;DR Why doesn’t Acclerate/FSDP seem to be doing much of anything to reduce memory in the following?
\nI’m trying to get some hands-on and learn how to run large models across multiple nodes and/or GPUs. I’m starting with Trainer/accelerate/FSDP2 and planning to work up from there but I think I’m missing something.
\npython 3.12.9
\ntorch 2.7.0
\ntransformers 4.52.4
\naccelerate 1.7.0
My “toy” program to train an “empty” model:
\nfrom datasets import Dataset, DatasetDict\nfrom transformers import AutoTokenizer, AutoConfig, AutoModelForCausalLM\n\nfrom transformers import DefaultDataCollator, DataCollatorForLanguageModeling\nfrom transformers import TrainingArguments, Trainer\nimport os\n\nmodel_dir = \'NousResearch/Llama-3.2-1B\'\nTRACE = False\nN = 2048\ncontext_length = 64\nbatch_size = 64\n\ndef load_datasets() :\n train_data_list = [\n {""text"" : ""The quick brown fox jumped over the lazy dog\'s back t{:06d}"".format(i)} for i in range(4*N)\n ]\n eval_data_list = [\n {""text"" : ""The quick brown fox jumped over the lazy dog\'s back e{:06d}"".format(i)} for i in range(N)\n ]\n datasets = DatasetDict ( # create datasets dict train and eval\n { \'train\': Dataset.from_list(train_data_list),\n \'eval\' : Dataset.from_list(eval_data_list)}\n )\n return datasets\n\ndef load_tokenizer(model_dir) :\n tokenizer = AutoTokenizer.from_pretrained(model_dir)\n return tokenizer\n\ndef load_model(model_dir) :\n # get just the config from the pretrained directory\n config = AutoConfig.from_pretrained(model_dir)\n model = AutoModelForCausalLM.from_config(config)\n return model\n\ndef mytrain(model_dir) :\n\n def tokenize(dataset) :\n return tokenizer(dataset[\'text\'], padding=\'max_length\', max_length=context_length, return_length=True)\n\n ##\n raw_datasets = load_datasets()\n if TRACE : print(""dataset\\n"", raw_datasets)\n ##\n tokenizer = load_tokenizer(model_dir)\n if TRACE : print(""tokenizer\\n"", tokenizer)\n ##\n tokenizer.pad_token = tokenizer.eos_token\n tokenized_datasets = raw_datasets.map(\n tokenize, batched=True, remove_columns=raw_datasets[""train""].column_names)\n if TRACE : print(""tokenized_datasets\\n"", tokenized_datasets)\n ##\n data_collator = DataCollatorForLanguageModeling(tokenizer, mlm=False)\n if TRACE :\n example_collated = data_collator([tokenized_datasets[""train""][i] for i in range(3)])\n print(""example_collated\\n"", example_collated)\n ##\n training_args = TrainingArguments( # do this before model load for FSDP?\n output_dir=""outputs/"",\n per_device_train_batch_size=batch_size,\n per_device_eval_batch_size=batch_size,\n num_train_epochs=10,\n logging_strategy=""epoch"",\n eval_strategy=""epoch"",\n save_strategy=""no"",\n push_to_hub=False,\n disable_tqdm=True,\n deepspeed=None,\n )\n ##\n model = load_model(model_dir) # do the after TrainingArguments which sets up some stuff?\n if TRACE : print(""model\\n"", model)\n ##\n trainer = Trainer(\n model=model,\n args=training_args,\n train_dataset=tokenized_datasets[""train""],\n eval_dataset=tokenized_datasets[""eval""],\n processing_class=tokenizer,\n data_collator=data_collator,\n )\n trainer.train()\n\nfrom datasets.utils.logging import disable_progress_bar\nimport torch\nif __name__ == ""__main__"" :\n disable_progress_bar()\n mytrain(\n model_dir=model_dir\n )\n torch.distributed.destroy_process_group()\nI first run my test progam as simple python/pytorch; single GPU without accelerate.
\n[gpu2:training] CUDA_VISIBLE_DEVICES=0 python 05_acctest.py \n{\'loss\': 0.8924, \'grad_norm\': 0.8125, \'learning_rate\': 4.50390625e-05, \'epoch\': 1.0}\n{\'eval_loss\': 2.5442957878112793, \'eval_runtime\': 2.4496, \'eval_samples_per_second\': 836.064, \'eval_steps_per_second\': 13.063, \'epoch\': 1.0}\n{\'loss\': 0.6293, \'grad_norm\': 0.65234375, \'learning_rate\': 4.00390625e-05, \'epoch\': 2.0}\n{\'eval_loss\': 2.6600184440612793, \'eval_runtime\': 2.4495, \'eval_samples_per_second\': 836.094, \'eval_steps_per_second\': 13.064, \'epoch\': 2.0}\n .\n .\n .\n{\'loss\': 0.6061, \'grad_norm\': 0.4921875, \'learning_rate\': 3.90625e-08, \'epoch\': 10.0}\n{\'eval_loss\': 2.8240463733673096, \'eval_runtime\': 2.4496, \'eval_samples_per_second\': 836.055, \'eval_steps_per_second\': 13.063, \'epoch\': 10.0}\n{\'train_runtime\': 333.183, \'train_samples_per_second\': 245.871, \'train_steps_per_second\': 3.842, \'train_loss\': 0.6405227959156037, \'epoch\': 10.0}\nWhile it’s running I use nvidia-smi to look at the memory used
\n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| 0 N/A N/A 21181 C python 21372MiB |\n+-----------------------------------------------------------------------------------------+\nThat’s at least in the ball-park for what accelerate estimates:
\n[gpu2:training] accelerate estimate-memory NousResearch/Llama-3.2-1B\nLoading pretrained config for `NousResearch/Llama-3.2-1B` from `transformers`...\n┌────────────────────────────────────────────────────────┐\n│ Memory Usage for loading `NousResearch/Llama-3.2-1B` │\n├───────┬─────────────┬──────────┬───────────────────────┤\n│ dtype │Largest Layer│Total Size│ Training using Adam │\n├───────┼─────────────┼──────────┼───────────────────────┤\n│float32│ 1002.0 MB │ 4.6 GB │ 18.42 GB │\n│float16│ 501.0 MB │ 2.3 GB │ 9.21 GB │\n│ int8 │ 250.5 MB │ 1.15 GB │ N/A │\n│ int4 │ 125.25 MB │589.28 MB │ N/A │\n└───────┴─────────────┴──────────┴───────────────────────┘\nNext I use “accelerate config” to generate a config file for 2 GPUs using FSDP2. (mostly with default values)
\n[gpu2:training] cat 1n2gfsdp_defaults.yaml \ncompute_environment: LOCAL_MACHINE\ndebug: false\ndistributed_type: FSDP\ndowncast_bf16: \'no\'\nenable_cpu_affinity: false\nfsdp_config:\n fsdp_activation_checkpointing: false\n fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP\n fsdp_cpu_ram_efficient_loading: true\n fsdp_offload_params: false\n fsdp_reshard_after_forward: true\n fsdp_state_dict_type: FULL_STATE_DICT\n fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer\n fsdp_version: 2\nmachine_rank: 0\nmain_training_function: main\nmixed_precision: \'no\'\nnum_machines: 1\nnum_processes: 2\nrdzv_backend: static\nsame_network: true\ntpu_env: []\ntpu_use_cluster: false\ntpu_use_sudo: false\nuse_cpu: false\nUsing that file an running with accelerate…
\n[gpu2:training] CUDA_VISIBLE_DEVICES=0,1 accelerate launch --config_file 1n2gfsdp_defaults.yaml 05_acctest.py \n{\'loss\': 1.0797, \'grad_norm\': 0.6328125, \'learning_rate\': 4.5078125000000006e-05, \'epoch\': 1.0}\n{\'eval_loss\': 2.5193161964416504, \'eval_runtime\': 1.376, \'eval_samples_per_second\': 1488.383, \'eval_steps_per_second\': 11.628, \'epoch\': 1.0}\n{\'loss\': 0.6584, \'grad_norm\': 0.4609375, \'learning_rate\': 4.0078125e-05, \'epoch\': 2.0}\n{\'eval_loss\': 2.5891079902648926, \'eval_runtime\': 1.3771, \'eval_samples_per_second\': 1487.218, \'eval_steps_per_second\': 11.619, \'epoch\': 2.0}\n .\n .\n .\n{\'loss\': 0.6096, \'grad_norm\': 0.462890625, \'learning_rate\': 7.8125e-08, \'epoch\': 10.0}\n{\'eval_loss\': 2.754133462905884, \'eval_runtime\': 1.3776, \'eval_samples_per_second\': 1486.605, \'eval_steps_per_second\': 11.614, \'epoch\': 10.0}\n{\'train_runtime\': 178.9799, \'train_samples_per_second\': 457.705, \'train_steps_per_second\': 3.576, \'train_loss\': 0.6661747217178344, \'epoch\': 10.0}\n… nvidia-smi memory during the computation…
\n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| 0 N/A N/A 24421 C ...AI/training-4.52.4/bin/python 21384MiB |\n| 1 N/A N/A 24422 C ...AI/training-4.52.4/bin/python 21388MiB |\n+-----------------------------------------------------------------------------------------+\nNext a config file with 4 GPUs…
\n[gpu2:training] cat 1n4gfsdp_defaults.yaml \ncompute_environment: LOCAL_MACHINE\ndebug: false\ndistributed_type: FSDP\ndowncast_bf16: \'no\'\nenable_cpu_affinity: false\nfsdp_config:\n fsdp_activation_checkpointing: false\n fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP\n fsdp_cpu_ram_efficient_loading: true\n fsdp_offload_params: false\n fsdp_reshard_after_forward: true\n fsdp_state_dict_type: FULL_STATE_DICT\n fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer\n fsdp_version: 2\nmachine_rank: 0\nmain_training_function: main\nmixed_precision: \'no\'\nnum_machines: 1\nnum_processes: 4\nrdzv_backend: static\nsame_network: true\ntpu_env: []\ntpu_use_cluster: false\ntpu_use_sudo: false\nuse_cpu: false\n… execute using accelerate…
\n[gpu2:training] CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch --config_file 1n4gfsdp_defaults.yaml 05_acctest.py \n{\'loss\': 1.373, \'grad_norm\': 0.458984375, \'learning_rate\': 4.515625e-05, \'epoch\': 1.0}\n{\'eval_loss\': 2.402463912963867, \'eval_runtime\': 0.6972, \'eval_samples_per_second\': 2937.372, \'eval_steps_per_second\': 11.474, \'epoch\': 1.0}\n{\'loss\': 0.7474, \'grad_norm\': 0.435546875, \'learning_rate\': 4.0156250000000004e-05, \'epoch\': 2.0}\n{\'eval_loss\': 2.3128156661987305, \'eval_runtime\': 0.6946, \'eval_samples_per_second\': 2948.607, \'eval_steps_per_second\': 11.518, \'epoch\': 2.0}\n .\n .\n .\n{\'loss\': 0.6214, \'grad_norm\': 0.30078125, \'learning_rate\': 1.5625e-07, \'epoch\': 10.0}\n{\'eval_loss\': 2.432434320449829, \'eval_runtime\': 0.694, \'eval_samples_per_second\': 2950.801, \'eval_steps_per_second\': 11.527, \'epoch\': 10.0}\n{\'train_runtime\': 89.6101, \'train_samples_per_second\': 914.182, \'train_steps_per_second\': 3.571, \'train_loss\': 0.718875628709793, \'epoch\': 10.0}\n… nvidia-smi while executing…
\n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| 0 N/A N/A 25570 C ...AI/training-4.52.4/bin/python 20526MiB |\n| 1 N/A N/A 25571 C ...AI/training-4.52.4/bin/python 20146MiB |\n| 2 N/A N/A 25572 C ...AI/training-4.52.4/bin/python 20146MiB |\n| 3 N/A N/A 25573 C ...AI/training-4.52.4/bin/python 20146MiB |\n+-----------------------------------------------------------------------------------------+\nClearly something is happening; I’m getting a performance benefit from using more GPUs (almost linear!). But, I’m not seeing a substantial improvement in memory usage.
\n===============================================================
\nI did a similar experiment with bloom-3b just to see if it made any difference, and things still seem strange.
\n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| 0 N/A N/A 37058 C python 74748MiB |\n+-----------------------------------------------------------------------------------------+\n\n┌────────────────────────────────────────────────────┐\n│ Memory Usage for loading `bigscience/bloom-3b` │\n├───────┬─────────────┬──────────┬───────────────────┤\n│ dtype │Largest Layer│Total Size│Training using Adam│\n├───────┼─────────────┼──────────┼───────────────────┤\n│float32│ 2.39 GB │ 11.19 GB │ 44.74 GB │\n│float16│ 1.2 GB │ 5.59 GB │ 22.37 GB │\n│ int8 │ 612.5 MB │ 2.8 GB │ N/A │\n│ int4 │ 306.25 MB │ 1.4 GB │ N/A │\n└───────┴─────────────┴──────────┴───────────────────┘\n\n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| 0 N/A N/A 251138 C ...AI/training-4.52.4/bin/python 53922MiB |\n| 1 N/A N/A 251139 C ...AI/training-4.52.4/bin/python 53538MiB |\n| 2 N/A N/A 251140 C ...AI/training-4.52.4/bin/python 53538MiB |\n| 3 N/A N/A 251141 C ...AI/training-4.52.4/bin/python 53538MiB |\n+-----------------------------------------------------------------------------------------+\nI don’t really understand how multi-GPU environments work…
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-05T06:24:05.499Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 158039, 'topic_slug': 'not-seeing-memory-benefit-to-accelerate-fsdp2', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/pytorch/torchtitan/issues/735', 'internal': False, 'reflection': False, 'title': '[question]FSDP2 have more peak active memory/reserved memory than FSDP1 · Issue #735 · pytorch/torchtitan · GitHub', 'clicks': 6}, {'url': 'https://github.com/pytorch/torchtune/issues/2402', 'internal': False, 'reflection': False, 'title': 'Does FSDP v2 have the best performance? · Issue #2402 · pytorch/torchtune · GitHub', 'clicks': 5}, {'url': 'https://github.com/pytorch/pytorch/issues/147168', 'internal': False, 'reflection': False, 'title': '[FSDP2] The evil `record_stream` in c10d causes FSDP2 to over-allocate GPU memory · Issue #147168 · pytorch/pytorch · GitHub', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/not-seeing-memory-benefit-to-accelerate-fsdp2/158039/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228173, 'name': 'hpcpony', 'username': 'hpcpony', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/h/779978/{size}.png', 'created_at': '2025-06-18T15:49:22.924Z', 'cooked': 'So after much futzing around and doing FSDP from pytorch I discovered that the answer to this question is that the memory usage reported by nvidia-smi is not an accurate reflection of memory required/used by pytorch. Apparently pytorch maintains a cache which is greater than that needed/used and that is primarily what the nvidia number reflects.
\npytorch.cuda has a number of ways to get memory information that seems to be more relevant (though not always clear of the implications).
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-18T15:49:22.924Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 10, 'reads': 3, 'readers_count': 2, 'score': 65.6, 'yours': False, 'topic_id': 158039, 'topic_slug': 'not-seeing-memory-benefit-to-accelerate-fsdp2', 'display_username': 'hpcpony', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96043, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/not-seeing-memory-benefit-to-accelerate-fsdp2/158039/3', 'reactions': [{'id': 'clap', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 228257, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-19T03:50:18.068Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-19T03:50:18.068Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 158039, 'topic_slug': 'not-seeing-memory-benefit-to-accelerate-fsdp2', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/not-seeing-memory-benefit-to-accelerate-fsdp2/158039/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","TL;DR Why doesn’t Acclerate/FSDP seem to be doing much of anything to reduce memory in the following?
+I’m trying to get some hands-on and learn how to run large models across multiple nodes and/or GPUs. I’m starting with Trainer/accelerate/FSDP2 and planning to work up from there but I think I’m missing something.
+python 3.12.9
+torch 2.7.0
+transformers 4.52.4
+accelerate 1.7.0
My “toy” program to train an “empty” model:
+from datasets import Dataset, DatasetDict
+from transformers import AutoTokenizer, AutoConfig, AutoModelForCausalLM
+
+from transformers import DefaultDataCollator, DataCollatorForLanguageModeling
+from transformers import TrainingArguments, Trainer
+import os
+
+model_dir = 'NousResearch/Llama-3.2-1B'
+TRACE = False
+N = 2048
+context_length = 64
+batch_size = 64
+
+def load_datasets() :
+ train_data_list = [
+ {""text"" : ""The quick brown fox jumped over the lazy dog's back t{:06d}"".format(i)} for i in range(4*N)
+ ]
+ eval_data_list = [
+ {""text"" : ""The quick brown fox jumped over the lazy dog's back e{:06d}"".format(i)} for i in range(N)
+ ]
+ datasets = DatasetDict ( # create datasets dict train and eval
+ { 'train': Dataset.from_list(train_data_list),
+ 'eval' : Dataset.from_list(eval_data_list)}
+ )
+ return datasets
+
+def load_tokenizer(model_dir) :
+ tokenizer = AutoTokenizer.from_pretrained(model_dir)
+ return tokenizer
+
+def load_model(model_dir) :
+ # get just the config from the pretrained directory
+ config = AutoConfig.from_pretrained(model_dir)
+ model = AutoModelForCausalLM.from_config(config)
+ return model
+
+def mytrain(model_dir) :
+
+ def tokenize(dataset) :
+ return tokenizer(dataset['text'], padding='max_length', max_length=context_length, return_length=True)
+
+ ##
+ raw_datasets = load_datasets()
+ if TRACE : print(""dataset\n"", raw_datasets)
+ ##
+ tokenizer = load_tokenizer(model_dir)
+ if TRACE : print(""tokenizer\n"", tokenizer)
+ ##
+ tokenizer.pad_token = tokenizer.eos_token
+ tokenized_datasets = raw_datasets.map(
+ tokenize, batched=True, remove_columns=raw_datasets[""train""].column_names)
+ if TRACE : print(""tokenized_datasets\n"", tokenized_datasets)
+ ##
+ data_collator = DataCollatorForLanguageModeling(tokenizer, mlm=False)
+ if TRACE :
+ example_collated = data_collator([tokenized_datasets[""train""][i] for i in range(3)])
+ print(""example_collated\n"", example_collated)
+ ##
+ training_args = TrainingArguments( # do this before model load for FSDP?
+ output_dir=""outputs/"",
+ per_device_train_batch_size=batch_size,
+ per_device_eval_batch_size=batch_size,
+ num_train_epochs=10,
+ logging_strategy=""epoch"",
+ eval_strategy=""epoch"",
+ save_strategy=""no"",
+ push_to_hub=False,
+ disable_tqdm=True,
+ deepspeed=None,
+ )
+ ##
+ model = load_model(model_dir) # do the after TrainingArguments which sets up some stuff?
+ if TRACE : print(""model\n"", model)
+ ##
+ trainer = Trainer(
+ model=model,
+ args=training_args,
+ train_dataset=tokenized_datasets[""train""],
+ eval_dataset=tokenized_datasets[""eval""],
+ processing_class=tokenizer,
+ data_collator=data_collator,
+ )
+ trainer.train()
+
+from datasets.utils.logging import disable_progress_bar
+import torch
+if __name__ == ""__main__"" :
+ disable_progress_bar()
+ mytrain(
+ model_dir=model_dir
+ )
+ torch.distributed.destroy_process_group()
+I first run my test progam as simple python/pytorch; single GPU without accelerate.
+[gpu2:training] CUDA_VISIBLE_DEVICES=0 python 05_acctest.py
+{'loss': 0.8924, 'grad_norm': 0.8125, 'learning_rate': 4.50390625e-05, 'epoch': 1.0}
+{'eval_loss': 2.5442957878112793, 'eval_runtime': 2.4496, 'eval_samples_per_second': 836.064, 'eval_steps_per_second': 13.063, 'epoch': 1.0}
+{'loss': 0.6293, 'grad_norm': 0.65234375, 'learning_rate': 4.00390625e-05, 'epoch': 2.0}
+{'eval_loss': 2.6600184440612793, 'eval_runtime': 2.4495, 'eval_samples_per_second': 836.094, 'eval_steps_per_second': 13.064, 'epoch': 2.0}
+ .
+ .
+ .
+{'loss': 0.6061, 'grad_norm': 0.4921875, 'learning_rate': 3.90625e-08, 'epoch': 10.0}
+{'eval_loss': 2.8240463733673096, 'eval_runtime': 2.4496, 'eval_samples_per_second': 836.055, 'eval_steps_per_second': 13.063, 'epoch': 10.0}
+{'train_runtime': 333.183, 'train_samples_per_second': 245.871, 'train_steps_per_second': 3.842, 'train_loss': 0.6405227959156037, 'epoch': 10.0}
+While it’s running I use nvidia-smi to look at the memory used
++-----------------------------------------------------------------------------------------+
+| Processes: |
+| GPU GI CI PID Type Process name GPU Memory |
+| ID ID Usage |
+|=========================================================================================|
+| 0 N/A N/A 21181 C python 21372MiB |
++-----------------------------------------------------------------------------------------+
+That’s at least in the ball-park for what accelerate estimates:
+[gpu2:training] accelerate estimate-memory NousResearch/Llama-3.2-1B
+Loading pretrained config for `NousResearch/Llama-3.2-1B` from `transformers`...
+┌────────────────────────────────────────────────────────┐
+│ Memory Usage for loading `NousResearch/Llama-3.2-1B` │
+├───────┬─────────────┬──────────┬───────────────────────┤
+│ dtype │Largest Layer│Total Size│ Training using Adam │
+├───────┼─────────────┼──────────┼───────────────────────┤
+│float32│ 1002.0 MB │ 4.6 GB │ 18.42 GB │
+│float16│ 501.0 MB │ 2.3 GB │ 9.21 GB │
+│ int8 │ 250.5 MB │ 1.15 GB │ N/A │
+│ int4 │ 125.25 MB │589.28 MB │ N/A │
+└───────┴─────────────┴──────────┴───────────────────────┘
+Next I use “accelerate config” to generate a config file for 2 GPUs using FSDP2. (mostly with default values)
+[gpu2:training] cat 1n2gfsdp_defaults.yaml
+compute_environment: LOCAL_MACHINE
+debug: false
+distributed_type: FSDP
+downcast_bf16: 'no'
+enable_cpu_affinity: false
+fsdp_config:
+ fsdp_activation_checkpointing: false
+ fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+ fsdp_cpu_ram_efficient_loading: true
+ fsdp_offload_params: false
+ fsdp_reshard_after_forward: true
+ fsdp_state_dict_type: FULL_STATE_DICT
+ fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
+ fsdp_version: 2
+machine_rank: 0
+main_training_function: main
+mixed_precision: 'no'
+num_machines: 1
+num_processes: 2
+rdzv_backend: static
+same_network: true
+tpu_env: []
+tpu_use_cluster: false
+tpu_use_sudo: false
+use_cpu: false
+Using that file an running with accelerate…
+[gpu2:training] CUDA_VISIBLE_DEVICES=0,1 accelerate launch --config_file 1n2gfsdp_defaults.yaml 05_acctest.py
+{'loss': 1.0797, 'grad_norm': 0.6328125, 'learning_rate': 4.5078125000000006e-05, 'epoch': 1.0}
+{'eval_loss': 2.5193161964416504, 'eval_runtime': 1.376, 'eval_samples_per_second': 1488.383, 'eval_steps_per_second': 11.628, 'epoch': 1.0}
+{'loss': 0.6584, 'grad_norm': 0.4609375, 'learning_rate': 4.0078125e-05, 'epoch': 2.0}
+{'eval_loss': 2.5891079902648926, 'eval_runtime': 1.3771, 'eval_samples_per_second': 1487.218, 'eval_steps_per_second': 11.619, 'epoch': 2.0}
+ .
+ .
+ .
+{'loss': 0.6096, 'grad_norm': 0.462890625, 'learning_rate': 7.8125e-08, 'epoch': 10.0}
+{'eval_loss': 2.754133462905884, 'eval_runtime': 1.3776, 'eval_samples_per_second': 1486.605, 'eval_steps_per_second': 11.614, 'epoch': 10.0}
+{'train_runtime': 178.9799, 'train_samples_per_second': 457.705, 'train_steps_per_second': 3.576, 'train_loss': 0.6661747217178344, 'epoch': 10.0}
+… nvidia-smi memory during the computation…
++-----------------------------------------------------------------------------------------+
+| Processes: |
+| GPU GI CI PID Type Process name GPU Memory |
+| ID ID Usage |
+|=========================================================================================|
+| 0 N/A N/A 24421 C ...AI/training-4.52.4/bin/python 21384MiB |
+| 1 N/A N/A 24422 C ...AI/training-4.52.4/bin/python 21388MiB |
++-----------------------------------------------------------------------------------------+
+Next a config file with 4 GPUs…
+[gpu2:training] cat 1n4gfsdp_defaults.yaml
+compute_environment: LOCAL_MACHINE
+debug: false
+distributed_type: FSDP
+downcast_bf16: 'no'
+enable_cpu_affinity: false
+fsdp_config:
+ fsdp_activation_checkpointing: false
+ fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+ fsdp_cpu_ram_efficient_loading: true
+ fsdp_offload_params: false
+ fsdp_reshard_after_forward: true
+ fsdp_state_dict_type: FULL_STATE_DICT
+ fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
+ fsdp_version: 2
+machine_rank: 0
+main_training_function: main
+mixed_precision: 'no'
+num_machines: 1
+num_processes: 4
+rdzv_backend: static
+same_network: true
+tpu_env: []
+tpu_use_cluster: false
+tpu_use_sudo: false
+use_cpu: false
+… execute using accelerate…
+[gpu2:training] CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch --config_file 1n4gfsdp_defaults.yaml 05_acctest.py
+{'loss': 1.373, 'grad_norm': 0.458984375, 'learning_rate': 4.515625e-05, 'epoch': 1.0}
+{'eval_loss': 2.402463912963867, 'eval_runtime': 0.6972, 'eval_samples_per_second': 2937.372, 'eval_steps_per_second': 11.474, 'epoch': 1.0}
+{'loss': 0.7474, 'grad_norm': 0.435546875, 'learning_rate': 4.0156250000000004e-05, 'epoch': 2.0}
+{'eval_loss': 2.3128156661987305, 'eval_runtime': 0.6946, 'eval_samples_per_second': 2948.607, 'eval_steps_per_second': 11.518, 'epoch': 2.0}
+ .
+ .
+ .
+{'loss': 0.6214, 'grad_norm': 0.30078125, 'learning_rate': 1.5625e-07, 'epoch': 10.0}
+{'eval_loss': 2.432434320449829, 'eval_runtime': 0.694, 'eval_samples_per_second': 2950.801, 'eval_steps_per_second': 11.527, 'epoch': 10.0}
+{'train_runtime': 89.6101, 'train_samples_per_second': 914.182, 'train_steps_per_second': 3.571, 'train_loss': 0.718875628709793, 'epoch': 10.0}
+… nvidia-smi while executing…
++-----------------------------------------------------------------------------------------+
+| Processes: |
+| GPU GI CI PID Type Process name GPU Memory |
+| ID ID Usage |
+|=========================================================================================|
+| 0 N/A N/A 25570 C ...AI/training-4.52.4/bin/python 20526MiB |
+| 1 N/A N/A 25571 C ...AI/training-4.52.4/bin/python 20146MiB |
+| 2 N/A N/A 25572 C ...AI/training-4.52.4/bin/python 20146MiB |
+| 3 N/A N/A 25573 C ...AI/training-4.52.4/bin/python 20146MiB |
++-----------------------------------------------------------------------------------------+
+Clearly something is happening; I’m getting a performance benefit from using more GPUs (almost linear!). But, I’m not seeing a substantial improvement in memory usage.
+===============================================================
+I did a similar experiment with bloom-3b just to see if it made any difference, and things still seem strange.
++-----------------------------------------------------------------------------------------+
+| Processes: |
+| GPU GI CI PID Type Process name GPU Memory |
+| ID ID Usage |
+|=========================================================================================|
+| 0 N/A N/A 37058 C python 74748MiB |
++-----------------------------------------------------------------------------------------+
+
+┌─────────────────────────────────────────���──────────┐
+│ Memory Usage for loading `bigscience/bloom-3b` │
+├───────┬─────────────┬──────────┬───────────────────┤
+│ dtype │Largest Layer│Total Size│Training using Adam│
+├───────┼─────────────┼──────────┼───────────────────┤
+│float32│ 2.39 GB │ 11.19 GB │ 44.74 GB │
+│float16│ 1.2 GB │ 5.59 GB │ 22.37 GB │
+│ int8 │ 612.5 MB │ 2.8 GB │ N/A │
+│ int4 │ 306.25 MB │ 1.4 GB │ N/A │
+└───────┴─────────────┴──────────┴───────────────────┘
+
++-----------------------------------------------------------------------------------------+
+| Processes: |
+| GPU GI CI PID Type Process name GPU Memory |
+| ID ID Usage |
+|=========================================================================================|
+| 0 N/A N/A 251138 C ...AI/training-4.52.4/bin/python 53922MiB |
+| 1 N/A N/A 251139 C ...AI/training-4.52.4/bin/python 53538MiB |
+| 2 N/A N/A 251140 C ...AI/training-4.52.4/bin/python 53538MiB |
+| 3 N/A N/A 251141 C ...AI/training-4.52.4/bin/python 53538MiB |
++-----------------------------------------------------------------------------------------+
+So after much futzing around and doing FSDP from pytorch I discovered that the answer to this question is that the memory usage reported by nvidia-smi is not an accurate reflection of memory required/used by pytorch. Apparently pytorch maintains a cache which is greater than that needed/used and that is primarily what the nvidia number reflects.
+pytorch.cuda has a number of ways to get memory information that seems to be more relevant (though not always clear of the implications).
" +Pytorch-Image models,https://discuss.huggingface.co/t/pytorch-image-models/154385,154385,13,2025-05-10 04:41:31.114000+00:00,"[{'id': 220959, 'name': 'Mohit Kumar', 'username': 'mohitb1i', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/dbc845/{size}.png', 'created_at': '2025-05-10T04:41:31.171Z', 'cooked': 'In the VisionTransformer class, the default act_layer is None . If we do not provide it - this will lead to a TypeError in MLP because none of the classes (Block , MLP , or VisionTransformer ) handle this case. Obvious error message:
\nTypeError: ‘NoneType’ object is not callable
Fix:
\nAlways set act_layer to a valid activation function (e.g., nn.GELU, nn.ReLU) when instantiating VisionTransformer.
\nExample:
import torch.nn as nn
\nmodel = VisionTransformer(act_layer=nn.GELU)
If not set, you’ll get TypeError: ‘NoneType’ object is not callable.
\nSolution provided by Triskel Data Deterministic AI.
', 'post_number': 2, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-06-10T20:24:42.368Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 10, 'readers_count': 9, 'score': 22.0, 'yours': False, 'topic_id': 154385, 'topic_slug': 'pytorch-image-models', 'display_username': 'Andrew Scott', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96276, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pytorch-image-models/154385/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226852, 'name': 'Daniela Brenes', 'username': 'dbrenes', 'avatar_template': '/user_avatar/discuss.huggingface.co/dbrenes/{size}/47087_2.png', 'created_at': '2025-06-11T00:05:50.417Z', 'cooked': 'Hello @mohitb1i ,
\nIn which PyTorch version are you experiencing this error?
\nMachine Learning Engineer at RidgeRun.ai
\nContact us: support@ridgerun.ai
I understand, but I am saying the default value of act_layer should be nn.GELU or just set it in instantiation, like:
\nblock_fn(\n...\nact_layer = act_layer or nn.GELU,\n...\n)\nNo it is a vision-transformer code from hugging face,
\noriginal repo
Upon reviewing the code, it appears that this behavior likely stems from the fact that the VisionTransformer class is not meant to be instantiated directly. Instead, the recommended approach is to use the timm.create_model function, which handles proper initialization of the available Vision Transformer variants. For example, calling models like vit_small_patch16_224 or vit_large_patch32_384 through timm.create_model returns a fully configured VisionTransformer instance.
However, if you choose to instantiate the VisionTransformer class directly, you are probably responsible for explicitly providing certain arguments—such as the act_layer—as you noted earlier.
Machine Learning Engineer at RidgeRun.ai
\nContact us: support@ridgerun.ai
import torch
\nimport torch.nn as nn
class VisionTransformer(nn.Module):
\ndef init(self, act_layer=None, **kwargs):
\nsuper().init()
\n# Default to GELU if none provided
\nif act_layer is None:
\nact_layer = nn.GELU
# Support both nn.ReLU and nn.ReLU() styles\n self.act = act_layer() if isinstance(act_layer, type) else act_layer\n\n # Example MLP block using activation\n self.mlp = nn.Sequential(\n nn.Linear(768, 3072),\n self.act,\n nn.Linear(3072, 768)\n )\n\ndef forward(self, x):\n return self.mlp(x)\nif name == “main”:
\nmodel = VisionTransformer()
\nx = torch.randn(1, 768)
\nout = model(x)
\nprint(out.shape)
Solution provided by Triskel Data Deterministic AI.
', 'post_number': 7, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-06-17T06:03:42.316Z', 'reply_count': 1, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 154385, 'topic_slug': 'pytorch-image-models', 'display_username': 'Andrew Scott', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 93201, 'username': 'dbrenes', 'name': 'Daniela Brenes', 'avatar_template': '/user_avatar/discuss.huggingface.co/dbrenes/{size}/47087_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96276, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pytorch-image-models/154385/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 228015, 'name': 'Mohit Kumar', 'username': 'mohitb1i', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/dbc845/{size}.png', 'created_at': '2025-06-17T19:12:21.511Z', 'cooked': 'Thanks, it was an oversight.
', 'post_number': 8, 'post_type': 1, 'posts_count': 9, 'updated_at': '2025-06-17T19:12:21.511Z', 'reply_count': 0, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 154385, 'topic_slug': 'pytorch-image-models', 'display_username': 'Mohit Kumar', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 96276, 'username': 'Pimpcat-AU', 'name': 'Andrew Scott', 'avatar_template': '/user_avatar/discuss.huggingface.co/pimpcat-au/{size}/48989_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93474, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/pytorch-image-models/154385/8', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 228108, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-18T07:12:51.633Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 9, 'post_type': 3, 'posts_count': 9, 'updated_at': '2025-06-18T07:12:51.633Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 154385, 'topic_slug': 'pytorch-image-models', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/pytorch-image-models/154385/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","In the VisionTransformer class, the default act_layer is None . If we do not provide it - this will lead to a TypeError in MLP because none of the classes (Block , MLP , or VisionTransformer ) handle this case. Obvious error message:
+TypeError: ‘NoneType’ object is not callable
import torch
+import torch.nn as nn
class VisionTransformer(nn.Module):
+def init(self, act_layer=None, **kwargs):
+super().init()
+# Default to GELU if none provided
+if act_layer is None:
+act_layer = nn.GELU
# Support both nn.ReLU and nn.ReLU() styles
+ self.act = act_layer() if isinstance(act_layer, type) else act_layer
+
+ # Example MLP block using activation
+ self.mlp = nn.Sequential(
+ nn.Linear(768, 3072),
+ self.act,
+ nn.Linear(3072, 768)
+ )
+
+def forward(self, x):
+ return self.mlp(x)
+if name == “main”:
+model = VisionTransformer()
+x = torch.randn(1, 768)
+out = model(x)
+print(out.shape)
Solution provided by Triskel Data Deterministic AI.
" +Cannot get tools to work: InferenceClient + hf-inference + Qwen/Qwen3-235B-A22B – Internal Server Error,https://discuss.huggingface.co/t/cannot-get-tools-to-work-inferenceclient-hf-inference-qwen-qwen3-235b-a22b-internal-server-error/159469,159469,6,2025-06-16 08:34:20.199000+00:00,"[{'id': 227679, 'name': 'Björn Buchhold', 'username': 'bbuchhold', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/c2a13f/{size}.png', 'created_at': '2025-06-16T08:34:20.253Z', 'cooked': 'I’m trying to get an existing app (OpenAI or Gemini both work well ) to run on open-weight models and keep failing. I have now distilled a minimal example that works on gpt-4.1-mini but doesn’t on Qwen3.
\nclient = openai.Client()\nMODEL = ""gpt-4.1-mini""\n\nmessages = [\n {""role"": ""user"", ""content"": ""You are a shopping assistant for a store. You can help pick the right products for the user.""},\n {""role"": ""user"", ""content"": ""I\'m looking for a T-shirt""}\n]\n\ndummy_tools = [{\n ""type"": ""function"",\n ""function"": {\n ""name"": ""get_products"",\n ""description"": (\n ""Search for products. Useful if someone needs clothing.""\n ),\n ""parameters"": {\n ""type"": ""object"",\n ""properties"": {\n ""query"": {\n ""type"": ""string"",\n ""description"": ""The query to look up products for.""\n }\n },\n ""required"": [\n ""query""\n ],\n ""additionalProperties"": False\n },\n ""strict"": True\n }\n }]\nr = client.chat.completions.create(model=MODEL, tools=dummy_tools, messages=messages)\ntcs = []\nfor tc in r.choices[0].message.tool_calls:\n tcs.append({\n ""id"": tc.id,\n ""type"": tc.type,\n ""function"": {\n ""name"": tc.function.name,\n ""arguments"": tc.function.arguments,\n }\n })\nmessages.append({""role"": ""assistant"", ""tool_calls"": tcs})\n# fake it for brevity\nmessages.append({""role"": ""tool"", ""tool_call_id"": tcs[0][""id""], ""content"": ""Product 1: Blue T-Shirt\\nProduct 2: Red Hoody.""})\nfor m in messages:\n print(m)\nprint(""-----------"")\nr = client.chat.completions.create(model=MODEL, messages=messages)\nprint(r.choices[0])\nworks and prints:
\n{\'role\': \'user\', \'content\': \'You are a shopping assistant for a store. You can help pick the right products for the user.\'}\n{\'role\': \'user\', \'content\': ""I\'m looking for a T-shirt""}\n{\'role\': \'assistant\', \'tool_calls\': [{\'id\': \'call_b7Gp98ZGcdv6TSbAlgrZC8Sq\', \'type\': \'function\', \'function\': {\'name\': \'get_products\', \'arguments\': \'{""query"":""T-shirt""}\'}}]}\n{\'role\': \'tool\', \'tool_call_id\': \'call_b7Gp98ZGcdv6TSbAlgrZC8Sq\', \'content\': \'Product 1: Blue T-Shirt\\nProduct 2: Red Hoody.\'}\n -----------\nChoice(finish_reason=\'stop\', index=0, logprobs=None, message=ChatCompletionMessage(content=\'I found a Blue T-Shirt for you. Would you like more options or details about this one?\', refusal=None, role=\'assistant\', annotations=[], audio=None, function_call=None, tool_calls=None))\nMeanwhile:
\nclient = InferenceClient(\n provider=""hf-inference"",\n api_key=os.environ[""HF_TOKEN""],\n )\nMODEL = ""Qwen/Qwen3-235B-A22B""\n\nmessages = [\n {""role"": ""user"", ""content"": ""You are a shopping assistant for a store. You can help pick the right products for the user.""},\n {""role"": ""user"", ""content"": ""I\'m looking for a T-shirt""}\n]\n\ndummy_tools = [{\n ""type"": ""function"",\n ""function"": {\n ""name"": ""get_products"",\n ""description"": (\n ""Search for products. Useful if someone needs clothing.""\n ),\n ""parameters"": {\n ""type"": ""object"",\n ""properties"": {\n ""query"": {\n ""type"": ""string"",\n ""description"": ""The query to look up products for.""\n }\n },\n ""required"": [\n ""query""\n ],\n ""additionalProperties"": False\n },\n ""strict"": True\n }\n }]\nr = client.chat.completions.create(model=MODEL, tools=dummy_tools, messages=messages)\ntcs = []\nfor tc in r.choices[0].message.tool_calls:\n tcs.append({\n ""id"": tc.id,\n ""type"": tc.type,\n ""function"": {\n ""name"": tc.function.name,\n ""arguments"": tc.function.arguments,\n }\n })\nmessages.append({""role"": ""assistant"", ""tool_calls"": tcs})\n# fake it for brevity\nmessages.append({""role"": ""tool"", ""tool_call_id"": tcs[0][""id""], ""content"": ""Product 1: Blue T-Shirt\\nProduct 2: Red Hoody.""})\nfor m in messages:\n print(m)\nprint(""-----------"")\nr = client.chat.completions.create(model=MODEL, messages=messages)\nprint(r.choices[0])\nfails with
\n---------------------------------------------------------------------------\nHTTPError Traceback (most recent call last)\nFile ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/utils/_http.py:409, in hf_raise_for_status(response, endpoint_name)\n 408 try:\n--> 409 response.raise_for_status()\n 410 except HTTPError as e:\n\nFile ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/requests/models.py:1024, in Response.raise_for_status(self)\n 1023 if http_error_msg:\n-> 1024 raise HTTPError(http_error_msg, response=self)\n\nHTTPError: 500 Server Error: Internal Server Error for url: https://router.huggingface.co/hf-inference/models/Qwen/Qwen3-235B-A22B/v1/chat/completions\n\nThe above exception was the direct cause of the following exception:\n\nHfHubHTTPError Traceback (most recent call last)\nCell In[107], line 52\n 50 print(m)\n 51 print(""-----------"")\n---> 52 r = client.chat.completions.create(model=MODEL, messages=messages)\n 53 print(r.choices[0])\n\nFile ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/inference/_client.py:924, in InferenceClient.chat_completion(self, messages, model, stream, frequency_penalty, logit_bias, logprobs, max_tokens, n, presence_penalty, response_format, seed, stop, stream_options, temperature, tool_choice, tool_prompt, tools, top_logprobs, top_p, extra_body)\n 896 parameters = {\n 897 ""model"": payload_model,\n 898 ""frequency_penalty"": frequency_penalty,\n (...) 915 **(extra_body or {}),\n 916 }\n 917 request_parameters = provider_helper.prepare_request(\n 918 inputs=messages,\n 919 parameters=parameters,\n (...) 922 api_key=self.token,\n 923 )\n--> 924 data = self._inner_post(request_parameters, stream=stream)\n 926 if stream:\n 927 return _stream_chat_completion_response(data) # type: ignore[arg-type]\n\nFile ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/inference/_client.py:280, in InferenceClient._inner_post(self, request_parameters, stream)\n 277 raise InferenceTimeoutError(f""Inference call timed out: {request_parameters.url}"") from error # type: ignore\n 279 try:\n--> 280 hf_raise_for_status(response)\n 281 return response.iter_lines() if stream else response.content\n 282 except HTTPError as error:\n\nFile ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/utils/_http.py:482, in hf_raise_for_status(response, endpoint_name)\n 478 raise _format(HfHubHTTPError, message, response) from e\n 480 # Convert `HTTPError` into a `HfHubHTTPError` to display request information\n 481 # as well (request id and/or server error message)\n--> 482 raise _format(HfHubHTTPError, str(e), response) from e\n\nHfHubHTTPError: 500 Server Error: Internal Server Error for url: https://router.huggingface.co/hf-inference/models/Qwen/Qwen3-235B-A22B/v1/chat/completions (Request ID: Root=1-684c0e94-1b2fcc1112ce97d968f42b89;4a0857fe-92d3-4b59-977c-2c58fee78502)\nUnfortunately, I fail to get a better reason than the 500 return code, and I’m not sure if I am misusing the API somehow
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-16T08:34:20.253Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 42, 'reads': 10, 'readers_count': 9, 'score': 217.0, 'yours': False, 'topic_id': 159469, 'topic_slug': 'cannot-get-tools-to-work-inferenceclient-hf-inference-qwen-qwen3-235b-a22b-internal-server-error', 'display_username': 'Björn Buchhold', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/bad-request-your-endpoint-is-in-error-check-its-status-on-endpoints-huggingface-co/159439/5', 'internal': True, 'reflection': True, 'title': '""Bad Request: Your endpoint is in error, check its status on endpoints.huggingface.co', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96853, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cannot-get-tools-to-work-inferenceclient-hf-inference-qwen-qwen3-235b-a22b-internal-server-error/159469/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 227702, 'name': 'Björn Buchhold', 'username': 'bbuchhold', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/b/c2a13f/{size}.png', 'created_at': '2025-06-16T08:56:17.694Z', 'cooked': '3 days later, this works. I assume the “internal server error” actually was an internal error after all
Great. Links that may be useful in case of trouble. However, ongoing problems may not always be apparent.
\nServer status: https://status.huggingface.co/
\nChangeLog: Changelog - Hugging Face
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-17T01:55:03.232Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 159469, 'topic_slug': 'cannot-get-tools-to-work-inferenceclient-hf-inference-qwen-qwen3-235b-a22b-internal-server-error', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cannot-get-tools-to-work-inferenceclient-hf-inference-qwen-qwen3-235b-a22b-internal-server-error/159469/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m trying to get an existing app (OpenAI or Gemini both work well ) to run on open-weight models and keep failing. I have now distilled a minimal example that works on gpt-4.1-mini but doesn’t on Qwen3.
+client = openai.Client()
+MODEL = ""gpt-4.1-mini""
+
+messages = [
+ {""role"": ""user"", ""content"": ""You are a shopping assistant for a store. You can help pick the right products for the user.""},
+ {""role"": ""user"", ""content"": ""I'm looking for a T-shirt""}
+]
+
+dummy_tools = [{
+ ""type"": ""function"",
+ ""function"": {
+ ""name"": ""get_products"",
+ ""description"": (
+ ""Search for products. Useful if someone needs clothing.""
+ ),
+ ""parameters"": {
+ ""type"": ""object"",
+ ""properties"": {
+ ""query"": {
+ ""type"": ""string"",
+ ""description"": ""The query to look up products for.""
+ }
+ },
+ ""required"": [
+ ""query""
+ ],
+ ""additionalProperties"": False
+ },
+ ""strict"": True
+ }
+ }]
+r = client.chat.completions.create(model=MODEL, tools=dummy_tools, messages=messages)
+tcs = []
+for tc in r.choices[0].message.tool_calls:
+ tcs.append({
+ ""id"": tc.id,
+ ""type"": tc.type,
+ ""function"": {
+ ""name"": tc.function.name,
+ ""arguments"": tc.function.arguments,
+ }
+ })
+messages.append({""role"": ""assistant"", ""tool_calls"": tcs})
+# fake it for brevity
+messages.append({""role"": ""tool"", ""tool_call_id"": tcs[0][""id""], ""content"": ""Product 1: Blue T-Shirt\nProduct 2: Red Hoody.""})
+for m in messages:
+ print(m)
+print(""-----------"")
+r = client.chat.completions.create(model=MODEL, messages=messages)
+print(r.choices[0])
+works and prints:
+{'role': 'user', 'content': 'You are a shopping assistant for a store. You can help pick the right products for the user.'}
+{'role': 'user', 'content': ""I'm looking for a T-shirt""}
+{'role': 'assistant', 'tool_calls': [{'id': 'call_b7Gp98ZGcdv6TSbAlgrZC8Sq', 'type': 'function', 'function': {'name': 'get_products', 'arguments': '{""query"":""T-shirt""}'}}]}
+{'role': 'tool', 'tool_call_id': 'call_b7Gp98ZGcdv6TSbAlgrZC8Sq', 'content': 'Product 1: Blue T-Shirt\nProduct 2: Red Hoody.'}
+ -----------
+Choice(finish_reason='stop', index=0, logprobs=None, message=ChatCompletionMessage(content='I found a Blue T-Shirt for you. Would you like more options or details about this one?', refusal=None, role='assistant', annotations=[], audio=None, function_call=None, tool_calls=None))
+Meanwhile:
+client = InferenceClient(
+ provider=""hf-inference"",
+ api_key=os.environ[""HF_TOKEN""],
+ )
+MODEL = ""Qwen/Qwen3-235B-A22B""
+
+messages = [
+ {""role"": ""user"", ""content"": ""You are a shopping assistant for a store. You can help pick the right products for the user.""},
+ {""role"": ""user"", ""content"": ""I'm looking for a T-shirt""}
+]
+
+dummy_tools = [{
+ ""type"": ""function"",
+ ""function"": {
+ ""name"": ""get_products"",
+ ""description"": (
+ ""Search for products. Useful if someone needs clothing.""
+ ),
+ ""parameters"": {
+ ""type"": ""object"",
+ ""properties"": {
+ ""query"": {
+ ""type"": ""string"",
+ ""description"": ""The query to look up products for.""
+ }
+ },
+ ""required"": [
+ ""query""
+ ],
+ ""additionalProperties"": False
+ },
+ ""strict"": True
+ }
+ }]
+r = client.chat.completions.create(model=MODEL, tools=dummy_tools, messages=messages)
+tcs = []
+for tc in r.choices[0].message.tool_calls:
+ tcs.append({
+ ""id"": tc.id,
+ ""type"": tc.type,
+ ""function"": {
+ ""name"": tc.function.name,
+ ""arguments"": tc.function.arguments,
+ }
+ })
+messages.append({""role"": ""assistant"", ""tool_calls"": tcs})
+# fake it for brevity
+messages.append({""role"": ""tool"", ""tool_call_id"": tcs[0][""id""], ""content"": ""Product 1: Blue T-Shirt\nProduct 2: Red Hoody.""})
+for m in messages:
+ print(m)
+print(""-----------"")
+r = client.chat.completions.create(model=MODEL, messages=messages)
+print(r.choices[0])
+fails with
+---------------------------------------------------------------------------
+HTTPError Traceback (most recent call last)
+File ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/utils/_http.py:409, in hf_raise_for_status(response, endpoint_name)
+ 408 try:
+--> 409 response.raise_for_status()
+ 410 except HTTPError as e:
+
+File ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/requests/models.py:1024, in Response.raise_for_status(self)
+ 1023 if http_error_msg:
+-> 1024 raise HTTPError(http_error_msg, response=self)
+
+HTTPError: 500 Server Error: Internal Server Error for url: https://router.huggingface.co/hf-inference/models/Qwen/Qwen3-235B-A22B/v1/chat/completions
+
+The above exception was the direct cause of the following exception:
+
+HfHubHTTPError Traceback (most recent call last)
+Cell In[107], line 52
+ 50 print(m)
+ 51 print(""-----------"")
+---> 52 r = client.chat.completions.create(model=MODEL, messages=messages)
+ 53 print(r.choices[0])
+
+File ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/inference/_client.py:924, in InferenceClient.chat_completion(self, messages, model, stream, frequency_penalty, logit_bias, logprobs, max_tokens, n, presence_penalty, response_format, seed, stop, stream_options, temperature, tool_choice, tool_prompt, tools, top_logprobs, top_p, extra_body)
+ 896 parameters = {
+ 897 ""model"": payload_model,
+ 898 ""frequency_penalty"": frequency_penalty,
+ (...) 915 **(extra_body or {}),
+ 916 }
+ 917 request_parameters = provider_helper.prepare_request(
+ 918 inputs=messages,
+ 919 parameters=parameters,
+ (...) 922 api_key=self.token,
+ 923 )
+--> 924 data = self._inner_post(request_parameters, stream=stream)
+ 926 if stream:
+ 927 return _stream_chat_completion_response(data) # type: ignore[arg-type]
+
+File ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/inference/_client.py:280, in InferenceClient._inner_post(self, request_parameters, stream)
+ 277 raise InferenceTimeoutError(f""Inference call timed out: {request_parameters.url}"") from error # type: ignore
+ 279 try:
+--> 280 hf_raise_for_status(response)
+ 281 return response.iter_lines() if stream else response.content
+ 282 except HTTPError as error:
+
+File ~/micromamba/envs/strauss_rag_202505/lib/python3.13/site-packages/huggingface_hub/utils/_http.py:482, in hf_raise_for_status(response, endpoint_name)
+ 478 raise _format(HfHubHTTPError, message, response) from e
+ 480 # Convert `HTTPError` into a `HfHubHTTPError` to display request information
+ 481 # as well (request id and/or server error message)
+--> 482 raise _format(HfHubHTTPError, str(e), response) from e
+
+HfHubHTTPError: 500 Server Error: Internal Server Error for url: https://router.huggingface.co/hf-inference/models/Qwen/Qwen3-235B-A22B/v1/chat/completions (Request ID: Root=1-684c0e94-1b2fcc1112ce97d968f42b89;4a0857fe-92d3-4b59-977c-2c58fee78502)
+Unfortunately, I fail to get a better reason than the 500 return code, and I’m not sure if I am misusing the API somehow
","3 days later, this works. I assume the “internal server error” actually was an internal error after all
Hello everyone,
\nI am trying to fine-tune a Llama 3.1 8B Instruct Model using LoRA. I would like to use multiple GPUs, but I am getting the following error.
Traceback (most recent call last): \n File ""/home/user/s25/finetune_model_LoRA.py"", line 68, in <module> \n trainer.train() \n ~~~~~~~~~~~~~^^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 2240, in train \n return inner_training_loop( \n args=args, \n ...<2 lines>... \n ignore_keys_for_eval=ignore_keys_for_eval, \n ) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 2555, in _inner_training_loop \n tr_loss_step = self.training_step(model, inputs, num_items_in_batch) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/trl/trainer/sft_trainer.py"", line 733, in training_step \n return super().training_step(*args, **kwargs) \n ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 3745, in training_step \n loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/trl/trainer/sft_trainer.py"", line 687, in compute_loss \n (loss, outputs) = super().compute_loss( \n ~~~~~~~~~~~~~~~~~~~~^ \n model, inputs, return_outputs=True, num_items_in_batch=num_items_in_batch \n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ \n ) \n ^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 3810, in compute_loss \n outputs = model(**inputs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1751, in _wrapped_call_impl \n return self._call_impl(*args, **kwargs) \n ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1762, in _call_impl \n return forward_call(*args, **kwargs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/accelerate/utils/operations.py"", line 818, in forward \n return model_forward(*args, **kwargs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/accelerate/utils/operations.py"", line 806, in __call__ \n return convert_to_fp32(self.model_forward(*args, **kwargs)) \n ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/amp/autocast_mode.py"", line 44, in decorate_autocast \n return func(*args, **kwargs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/peft/peft_model.py"", line 1757, in forward \n return self.base_model( \n ~~~~~~~~~~~~~~~^ \n input_ids=input_ids, \n ^^^^^^^^^^^^^^^^^^^^ \n ...<6 lines>... \n **kwargs, \n ^^^^^^^^^ \n ) \n ^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1751, in _wrapped_call_impl \n return self._call_impl(*args, **kwargs) \n ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1762, in _call_impl \n return forward_call(*args, **kwargs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/peft/tuners/tuners_utils.py"", line 193, in forward \n return self.model.forward(*args, **kwargs) \n ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/accelerate/hooks.py"", line 175, in new_forward \n output = module._old_forward(*args, **kwargs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/utils/generic.py"", line 969, in wrapper\n output = func(self, *args, **kwargs) \n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/models/llama/modeling_llama.py"", line 708, in forward\n loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)\n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/loss/loss_utils.py"", line 64, in ForCausalLMLoss\n loss = fixed_cross_entropy(logits, shift_labels, num_items_in_batch, ignore_index, **kwargs)\n File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/loss/loss_utils.py"", line 38, in fixed_cross_entropy\n loss = loss / num_items_in_batch \n ~~~~~^~~~~~~~~~~~~~~~~~~~ \nRuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cuda:0!\nI use the following script.
\n# Imports\nfrom transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, DataCollatorForLanguageModeling, BitsAndBytesConfig\nfrom peft import LoraConfig\nfrom huggingface_hub import login\nfrom datasets import load_dataset\nfrom dotenv import load_dotenv\nfrom trl import SFTTrainer, SFTConfig\nfrom os import getenv\nimport torch\n\n# Load environment variables\nload_dotenv()\n\n# Login to huggingface\nlogin(token=getenv(""HUGGINGFACE_ACCESS_TOKEN""))\n\n# Load bitsandbytes config\nbnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type=""nf4"",\n bnb_4bit_compute_dtype=""float16"", bnb_4bit_use_double_quant=False)\n\n# Load the model and tokenizer corresponding to the model\nmodel_name = ""meta-llama/Llama-3.1-8B-Instruct""\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name, quantization_config=bnb_config, device_map=""auto"")\ntokenizer = AutoTokenizer.from_pretrained(model_name)\ntokenizer.pad_token = tokenizer.eos_token\n\n# Load the dataset\ndataset = load_dataset(\n ""json"", data_files=""/home/user/s25/documents.jsonl"", split=""train"")\n\n# Define tokenization function and tokenize the dataset\n\n\ndef tokenize(examples):\n inputs = tokenizer(examples[""document""])\n return inputs\n\n\ntokenized_dataset = dataset.map(\n tokenize, batched=True, remove_columns=[""document""])\n\n# Instantiate data collator\ndata_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)\n\n# Load LoRA configuration\npeft_config = LoraConfig(\n r=64, lora_alpha=16, lora_dropout=0, task_type=""CAUSAL_LM"", target_modules=[""q_proj"", ""k_proj"", ""v_proj"", ""o_proj"", ""gate_proj"", ""up_proj"", ""down_proj""])\n\n# Specify the training arguments\ntrainings_arguments = SFTConfig(output_dir=""/data/projects/s25/Llama-3.1-8B-Instruct-lora-v6-1epochs"", save_strategy=""steps"", save_steps=500, save_total_limit=1,\n per_device_train_batch_size=2, num_train_epochs=1, learning_rate=5e-4, weight_decay=0.01, logging_dir=""/data/projects/s25/Llama-3.1-8B-Instruct-lora-v6-1epochs-log"", logging_steps=50, report_to=""none"", fp16=True, bf16=False, dataset_text_field=None)\n\n# Set up trainer\ntrainer = SFTTrainer(model=model, args=trainings_arguments,\n train_dataset=tokenized_dataset, processing_class=tokenizer, data_collator=data_collator, peft_config=peft_config)\n\n# Train the model\ntrainer.train()\nThis issue is very similar to the following already existing posts:
\n\n\nHowever, the solutions provided there did not help me solve the problem.
\nLastly, the versions of the most relevant packages (not necessarily enough to run the script, but I was character-limited for this post).
\naccelerate 1.7.0 pyhe01879c_0 conda-forge \nbitsandbytes 0.46.0 cuda126_py313hde49398_0 conda-forge \ndatasets 3.6.0 pyhd8ed1ab_0 conda-forge\nhuggingface_hub 0.33.0 pyhd8ed1ab_0 conda-forge \nnumpy 2.3.0 py313h17eae1a_0 conda-forge \npandas 2.3.0 py313ha87cce1_0 conda-forge \npip 25.1.1 pyh145f28c_0 conda-forge \npython 3.13.2 hf636f53_101_cp313 conda-forge \npython-dateutil 2.9.0.post0 pyhff2d567_1 conda-forge \npython-dotenv 1.1.0 pyh29332c3_1 conda-forge \npython-gil 3.13.5 h4df99d1_101 conda-forge \npython-tzdata 2025.2 pyhd8ed1ab_0 conda-forge \npython-xxhash 3.5.0 py313h536fd9c_2 conda-forge \npython_abi 3.13 7_cp313 conda-forge \npytorch 2.7.0 cuda126_generic_py313_h14c909a_200 conda-forge \ntokenizers 0.21.1 py313h1191936_0 conda-forge\ntorch 2.6.0+cu126 pypi_0 pypi\ntorchaudio 2.6.0+cu126 pypi_0 pypi\ntorchvision 0.21.0+cu126 pypi_0 pypi\ntransformers 4.52.4 pyhd8ed1ab_0 conda-forge\ntrl 0.18.2 pyhd8ed1ab_0 conda-forge\nI am very grateful for any support! Thank you very much!
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-16T06:41:51.002Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 128, 'reads': 7, 'readers_count': 6, 'score': 586.4, 'yours': False, 'topic_id': 159445, 'topic_slug': 'lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0', 'display_username': 'Benjamin Koch', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-7-and-cuda-0/147337', 'internal': True, 'reflection': False, 'title': 'RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:7 and cuda:0!', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97059, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0/159445/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 227649, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-16T07:00:48.906Z', 'cooked': 'If so, it may be an unresolved compatibility issue between accelerate and bitsandbytes?
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-16T07:00:48.906Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 14, 'reads': 7, 'readers_count': 6, 'score': 66.4, 'yours': False, 'topic_id': 159445, 'topic_slug': 'lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/bitsandbytes-conflict-with-accelerate/150275', 'internal': True, 'reflection': False, 'title': 'BitsandBytes conflict with Accelerate', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/multi-gpu-inference-llama-3-2-vision-with-qlora/150685', 'internal': True, 'reflection': False, 'title': 'Multi-gpu inference llama-3.2 vision with QLoRA', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0/159445/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 227650, 'name': 'Benjamin Koch', 'username': 'by-benj-k', 'avatar_template': '/user_avatar/discuss.huggingface.co/by-benj-k/{size}/49508_2.png', 'created_at': '2025-06-16T07:22:17.905Z', 'cooked': 'Thanks for the information, however, I have tried running the script without the bitsandbytes configuration (and also with the bitsandbytes package removed) by just utilizing more GPUs, however the error seems to persist.
\nSo essentially by simply loading the model as follows:
\nmodel_name = ""meta-llama/Llama-3.1-8B-Instruct""\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name, device_map=""auto"")\ntokenizer = AutoTokenizer.from_pretrained(model_name)\ntokenizer.pad_token = tokenizer.eos_token\n(And by the way launching the script with: CUDA_VISIBLE_DEVICES=0,1 python finetune_model_LoRA.py)
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-16T07:26:23.606Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 6, 'readers_count': 5, 'score': 26.2, 'yours': False, 'topic_id': 159445, 'topic_slug': 'lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0', 'display_username': 'Benjamin Koch', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97059, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0/159445/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 227711, 'name': 'Benjamin Koch', 'username': 'by-benj-k', 'avatar_template': '/user_avatar/discuss.huggingface.co/by-benj-k/{size}/49508_2.png', 'created_at': '2025-06-16T09:44:18.325Z', 'cooked': 'UPDATE: at least for now the problem seems to be fixed. I downgraded the transformers library to version 4.49.0, used the transfomers.Trainer instead of the SFTTrainer and modified the loading of the model to the following.
\n# Load bitsandbytes config\nbnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type=""nf4"",\n bnb_4bit_compute_dtype=""float16"", bnb_4bit_use_double_quant=False)\n\n# Load LoRA configuration\npeft_config = LoraConfig(\n r=64, lora_alpha=16, lora_dropout=0, task_type=""CAUSAL_LM"", target_modules=[""q_proj"", ""k_proj"", ""v_proj"", ""o_proj"", ""gate_proj"", ""up_proj"", ""down_proj""])\n\n# Load the model and prepare it for peft finetuning\nmodel_name = ""meta-llama/Llama-3.1-8B-Instruct""\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name, quantization_config=bnb_config, device_map=""auto"")\n\nmodel = prepare_model_for_kbit_training(model)\nmodel = get_peft_model(model, peft_config)\nMaybe this will help someone in the future!
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-16T09:44:18.325Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 5, 'readers_count': 4, 'score': 41.0, 'yours': False, 'topic_id': 159445, 'topic_slug': 'lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0', 'display_username': 'Benjamin Koch', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 97059, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0/159445/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 227832, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-16T21:45:04.711Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-16T21:45:04.711Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 159445, 'topic_slug': 'lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/lora-finetuning-runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-1-and-cuda-0/159445/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone,
+I am trying to fine-tune a Llama 3.1 8B Instruct Model using LoRA. I would like to use multiple GPUs, but I am getting the following error.
Traceback (most recent call last):
+ File ""/home/user/s25/finetune_model_LoRA.py"", line 68, in <module>
+ trainer.train()
+ ~~~~~~~~~~~~~^^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 2240, in train
+ return inner_training_loop(
+ args=args,
+ ...<2 lines>...
+ ignore_keys_for_eval=ignore_keys_for_eval,
+ )
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 2555, in _inner_training_loop
+ tr_loss_step = self.training_step(model, inputs, num_items_in_batch)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/trl/trainer/sft_trainer.py"", line 733, in training_step
+ return super().training_step(*args, **kwargs)
+ ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 3745, in training_step
+ loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/trl/trainer/sft_trainer.py"", line 687, in compute_loss
+ (loss, outputs) = super().compute_loss(
+ ~~~~~~~~~~~~~~~~~~~~^
+ model, inputs, return_outputs=True, num_items_in_batch=num_items_in_batch
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+ )
+ ^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/trainer.py"", line 3810, in compute_loss
+ outputs = model(**inputs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1751, in _wrapped_call_impl
+ return self._call_impl(*args, **kwargs)
+ ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1762, in _call_impl
+ return forward_call(*args, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/accelerate/utils/operations.py"", line 818, in forward
+ return model_forward(*args, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/accelerate/utils/operations.py"", line 806, in __call__
+ return convert_to_fp32(self.model_forward(*args, **kwargs))
+ ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/amp/autocast_mode.py"", line 44, in decorate_autocast
+ return func(*args, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/peft/peft_model.py"", line 1757, in forward
+ return self.base_model(
+ ~~~~~~~~~~~~~~~^
+ input_ids=input_ids,
+ ^^^^^^^^^^^^^^^^^^^^
+ ...<6 lines>...
+ **kwargs,
+ ^^^^^^^^^
+ )
+ ^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1751, in _wrapped_call_impl
+ return self._call_impl(*args, **kwargs)
+ ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/torch/nn/modules/module.py"", line 1762, in _call_impl
+ return forward_call(*args, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/peft/tuners/tuners_utils.py"", line 193, in forward
+ return self.model.forward(*args, **kwargs)
+ ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/accelerate/hooks.py"", line 175, in new_forward
+ output = module._old_forward(*args, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/utils/generic.py"", line 969, in wrapper
+ output = func(self, *args, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/models/llama/modeling_llama.py"", line 708, in forward
+ loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/loss/loss_utils.py"", line 64, in ForCausalLMLoss
+ loss = fixed_cross_entropy(logits, shift_labels, num_items_in_batch, ignore_index, **kwargs)
+ File ""/local/home/user/miniforge3/envs/project/lib/python3.13/site-packages/transformers/loss/loss_utils.py"", line 38, in fixed_cross_entropy
+ loss = loss / num_items_in_batch
+ ~~~~~^~~~~~~~~~~~~~~~~~~~
+RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cuda:0!
+I use the following script.
+# Imports
+from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, DataCollatorForLanguageModeling, BitsAndBytesConfig
+from peft import LoraConfig
+from huggingface_hub import login
+from datasets import load_dataset
+from dotenv import load_dotenv
+from trl import SFTTrainer, SFTConfig
+from os import getenv
+import torch
+
+# Load environment variables
+load_dotenv()
+
+# Login to huggingface
+login(token=getenv(""HUGGINGFACE_ACCESS_TOKEN""))
+
+# Load bitsandbytes config
+bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_compute_dtype=""float16"", bnb_4bit_use_double_quant=False)
+
+# Load the model and tokenizer corresponding to the model
+model_name = ""meta-llama/Llama-3.1-8B-Instruct""
+model = AutoModelForCausalLM.from_pretrained(
+ model_name, quantization_config=bnb_config, device_map=""auto"")
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+tokenizer.pad_token = tokenizer.eos_token
+
+# Load the dataset
+dataset = load_dataset(
+ ""json"", data_files=""/home/user/s25/documents.jsonl"", split=""train"")
+
+# Define tokenization function and tokenize the dataset
+
+
+def tokenize(examples):
+ inputs = tokenizer(examples[""document""])
+ return inputs
+
+
+tokenized_dataset = dataset.map(
+ tokenize, batched=True, remove_columns=[""document""])
+
+# Instantiate data collator
+data_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)
+
+# Load LoRA configuration
+peft_config = LoraConfig(
+ r=64, lora_alpha=16, lora_dropout=0, task_type=""CAUSAL_LM"", target_modules=[""q_proj"", ""k_proj"", ""v_proj"", ""o_proj"", ""gate_proj"", ""up_proj"", ""down_proj""])
+
+# Specify the training arguments
+trainings_arguments = SFTConfig(output_dir=""/data/projects/s25/Llama-3.1-8B-Instruct-lora-v6-1epochs"", save_strategy=""steps"", save_steps=500, save_total_limit=1,
+ per_device_train_batch_size=2, num_train_epochs=1, learning_rate=5e-4, weight_decay=0.01, logging_dir=""/data/projects/s25/Llama-3.1-8B-Instruct-lora-v6-1epochs-log"", logging_steps=50, report_to=""none"", fp16=True, bf16=False, dataset_text_field=None)
+
+# Set up trainer
+trainer = SFTTrainer(model=model, args=trainings_arguments,
+ train_dataset=tokenized_dataset, processing_class=tokenizer, data_collator=data_collator, peft_config=peft_config)
+
+# Train the model
+trainer.train()
+This issue is very similar to the following already existing posts:
+ + +However, the solutions provided there did not help me solve the problem.
+Lastly, the versions of the most relevant packages (not necessarily enough to run the script, but I was character-limited for this post).
+accelerate 1.7.0 pyhe01879c_0 conda-forge
+bitsandbytes 0.46.0 cuda126_py313hde49398_0 conda-forge
+datasets 3.6.0 pyhd8ed1ab_0 conda-forge
+huggingface_hub 0.33.0 pyhd8ed1ab_0 conda-forge
+numpy 2.3.0 py313h17eae1a_0 conda-forge
+pandas 2.3.0 py313ha87cce1_0 conda-forge
+pip 25.1.1 pyh145f28c_0 conda-forge
+python 3.13.2 hf636f53_101_cp313 conda-forge
+python-dateutil 2.9.0.post0 pyhff2d567_1 conda-forge
+python-dotenv 1.1.0 pyh29332c3_1 conda-forge
+python-gil 3.13.5 h4df99d1_101 conda-forge
+python-tzdata 2025.2 pyhd8ed1ab_0 conda-forge
+python-xxhash 3.5.0 py313h536fd9c_2 conda-forge
+python_abi 3.13 7_cp313 conda-forge
+pytorch 2.7.0 cuda126_generic_py313_h14c909a_200 conda-forge
+tokenizers 0.21.1 py313h1191936_0 conda-forge
+torch 2.6.0+cu126 pypi_0 pypi
+torchaudio 2.6.0+cu126 pypi_0 pypi
+torchvision 0.21.0+cu126 pypi_0 pypi
+transformers 4.52.4 pyhd8ed1ab_0 conda-forge
+trl 0.18.2 pyhd8ed1ab_0 conda-forge
+I am very grateful for any support! Thank you very much!
","UPDATE: at least for now the problem seems to be fixed. I downgraded the transformers library to version 4.49.0, used the transfomers.Trainer instead of the SFTTrainer and modified the loading of the model to the following.
+# Load bitsandbytes config
+bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_compute_dtype=""float16"", bnb_4bit_use_double_quant=False)
+
+# Load LoRA configuration
+peft_config = LoraConfig(
+ r=64, lora_alpha=16, lora_dropout=0, task_type=""CAUSAL_LM"", target_modules=[""q_proj"", ""k_proj"", ""v_proj"", ""o_proj"", ""gate_proj"", ""up_proj"", ""down_proj""])
+
+# Load the model and prepare it for peft finetuning
+model_name = ""meta-llama/Llama-3.1-8B-Instruct""
+model = AutoModelForCausalLM.from_pretrained(
+ model_name, quantization_config=bnb_config, device_map=""auto"")
+
+model = prepare_model_for_kbit_training(model)
+model = get_peft_model(model, peft_config)
+Maybe this will help someone in the future!
" +"ValueError: Incompatible safetensors file. File metadata is not [‘pt’, ‘tf’, ‘flax’, ‘mlx’] but None",https://discuss.huggingface.co/t/valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none/159226,159226,13,2025-06-14 05:06:59.907000+00:00,"[{'id': 227369, 'name': 'Angkul', 'username': 'angkul07', 'avatar_template': '/user_avatar/discuss.huggingface.co/angkul07/{size}/49392_2.png', 'created_at': '2025-06-14T05:06:59.977Z', 'cooked': 'Hi experts,
\nI have trained a custom LLMs from scratch using pytorch and saved the model checkpoint. According to documentation, for custom pytorch models, I used the PyTorchModelHubMixin in my model class, to make it compatible. Now when I push it to hub using the following code:
GPT_CONFIG = {\n ""model_type"": ""gpt"",\n ""vocab_size"": 26000,\n ""context_length"": 256,\n ""emb_dim"": 768,\n ""n_heads"": 16,\n ""n_layers"": 12,\n ""drop_rate"": 0.2,\n ""qkv_bias"": False,\n ""flash"": True,\n}\n\nfrom model import GPTModel\nimport torch\n\nmodel = GPTModel(GPT_CONFIG)\n\ncheckpoint = torch.load(""/teamspace/studios/this_studio/model/gpt_model_checkpoint.pth"", map_location=""cpu"")\nmodel.load_state_dict(checkpoint[\'model_state_dict\'])\n\nmodel.save_pretrained(\n save_directory=""local-save-dir2"",\n config=GPT_CONFIG,\n)\n\nrepo_id = ""angkul07/llm_100M""\n\nmodel.push_to_hub(\n repo_id=repo_id,\n commit_message=""Initial commit of GPTModel checkpoint"",\n private=False\n)\nWhen I try to load it using the AutoModel:
model = AutoModel.from_pretrained(""angkul07/my-awesome-model"")\nI get the following Value error:
\nValueError: Incompatible safetensors file. File metadata is not [\'pt\', \'tf\', \'flax\', \'mlx\'] but None\n```.\n\n\nI have tried looking for it on the internet but its no help. So, how can I fix it? How can I add a metadata?This is a very rare error, but it may just be that there is no metadata.
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-14T07:13:18.284Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 26.6, 'yours': False, 'topic_id': 159226, 'topic_slug': 'valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/ml-explore/mlx/issues/743', 'internal': False, 'reflection': False, 'title': '[BUG] Saved safetensors are missing metadata format pt and cannot be loaded through `transformers` library · Issue #743 · ml-explore/mlx · GitHub', 'clicks': 15}, {'url': 'https://huggingface.co/SeaLLMs/SeaLLM-7B-Hybrid/discussions/2', 'internal': False, 'reflection': False, 'title': 'SeaLLMs/SeaLLM-7B-Hybrid · Seems like metadata is not in the safetensors files', 'clicks': 9}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none/159226/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 227383, 'name': 'Angkul', 'username': 'angkul07', 'avatar_template': '/user_avatar/discuss.huggingface.co/angkul07/{size}/49392_2.png', 'created_at': '2025-06-14T08:09:24.679Z', 'cooked': 'hey @John6666, thanks this works like a charm. Thank you so much.
\nBtw, I am facing one more issue, I have a custom trained sentencepiece tokenizer. So, two files tokenizer.model and tokenizer.vocab. Now, I want to convert them into the AutoTokenizer format to match the compatibility. I used the following code to convert:
from transformers import PreTrainedTokenizerFast\n\ntokenizer = PreTrainedTokenizerFast(\n tokenizer_file=""/teamspace/studios/this_studio/model/tokenizer.model"",\n model_max_length=256, \n bos_token=""<s>"",\n eos_token=""</s>"",\n unk_token=""<unk>"",\n pad_token=""<pad>"",\n mask_token=""<mask>"" \n)\n\ntokenizer.save_pretrained(""my-tokenizer"")\nBut I get the following error:
\nException: stream did not contain valid UTF-8\nDo you have any idea how to convert this sentencepiece tokenizer to AutoTokenizer format? Thanks.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-14T08:09:24.679Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 21.6, 'yours': False, 'topic_id': 159226, 'topic_slug': 'valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none', 'display_username': 'Angkul', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96913, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none/159226/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 227386, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-14T08:23:45.928Z', 'cooked': 'Maybe it’s a character encoding issue?
\nFor example, Windows 10 Notepad saves files in UTF-16, so comments that aren’t in English may cause errors…
\nThis probably won’t happen if you’re using VSCode, and if you’re using a Colab environment, the cause is likely something else.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-14T20:24:08.080Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 10.6, 'yours': False, 'topic_id': 159226, 'topic_slug': 'valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/valueerror-incompatible-safetensors-file-file-metadata-is-not-pt-tf-flax-mlx-but-none/159226/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi experts,
+I have trained a custom LLMs from scratch using pytorch and saved the model checkpoint. According to documentation, for custom pytorch models, I used the PyTorchModelHubMixin in my model class, to make it compatible. Now when I push it to hub using the following code:
GPT_CONFIG = {
+ ""model_type"": ""gpt"",
+ ""vocab_size"": 26000,
+ ""context_length"": 256,
+ ""emb_dim"": 768,
+ ""n_heads"": 16,
+ ""n_layers"": 12,
+ ""drop_rate"": 0.2,
+ ""qkv_bias"": False,
+ ""flash"": True,
+}
+
+from model import GPTModel
+import torch
+
+model = GPTModel(GPT_CONFIG)
+
+checkpoint = torch.load(""/teamspace/studios/this_studio/model/gpt_model_checkpoint.pth"", map_location=""cpu"")
+model.load_state_dict(checkpoint['model_state_dict'])
+
+model.save_pretrained(
+ save_directory=""local-save-dir2"",
+ config=GPT_CONFIG,
+)
+
+repo_id = ""angkul07/llm_100M""
+
+model.push_to_hub(
+ repo_id=repo_id,
+ commit_message=""Initial commit of GPTModel checkpoint"",
+ private=False
+)
+When I try to load it using the AutoModel:
model = AutoModel.from_pretrained(""angkul07/my-awesome-model"")
+I get the following Value error:
+ValueError: Incompatible safetensors file. File metadata is not ['pt', 'tf', 'flax', 'mlx'] but None
+```.
+
+
+I have tried looking for it on the internet but its no help. So, how can I fix it? How can I add a metadata?This is a very rare error, but it may just be that there is no metadata.
+ +" +Dataset.map Ignore failed batches,https://discuss.huggingface.co/t/dataset-map-ignore-failed-batches/158906,158906,10,2025-06-11 11:16:01.198000+00:00,"[{'id': 226940, 'name': 'wuwenhao', 'username': 'whh', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/w/958977/{size}.png', 'created_at': '2025-06-11T11:16:01.267Z', 'cooked': 'I often use the batch mode of dataset.map to process large amounts of data. Since there may be some format problems in the data, some batches may fail in the map (while most batches are OK).
\nIs there some way to ignore the failed batches and return the successful batches?
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-11T11:16:01.267Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 16, 'reads': 5, 'readers_count': 4, 'score': 96.0, 'yours': False, 'topic_id': 158906, 'topic_slug': 'dataset-map-ignore-failed-batches', 'display_username': 'wuwenhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 81967, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-map-ignore-failed-batches/158906/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 226948, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-11T11:39:10.983Z', 'cooked': 'For example, how about just use Python Exception?
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-11T11:39:10.983Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 158906, 'topic_slug': 'dataset-map-ignore-failed-batches', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/saving-outcomes-if-error-while-applying-map-function-on-dataset/31614', 'internal': True, 'reflection': False, 'title': 'Saving outcomes if Error while applying map function on dataset', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-map-ignore-failed-batches/158906/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 227235, 'name': 'wuwenhao', 'username': 'whh', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/w/958977/{size}.png', 'created_at': '2025-06-13T06:26:22.970Z', 'cooked': 'Thanks, It’s helpful !
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-13T06:26:22.970Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 158906, 'topic_slug': 'dataset-map-ignore-failed-batches', 'display_username': 'wuwenhao', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 81967, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/dataset-map-ignore-failed-batches/158906/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 227320, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-13T18:27:07.581Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-13T18:27:07.581Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 158906, 'topic_slug': 'dataset-map-ignore-failed-batches', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/dataset-map-ignore-failed-batches/158906/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I often use the batch mode of dataset.map to process large amounts of data. Since there may be some format problems in the data, some batches may fail in the map (while most batches are OK).
+Is there some way to ignore the failed batches and return the successful batches?
","For example, how about just use Python Exception?
+" +Unable to Upload arXiv Paper to HuggingFace Daily Papers,https://discuss.huggingface.co/t/unable-to-upload-arxiv-paper-to-huggingface-daily-papers/159000,159000,23,2025-06-12 02:21:34.885000+00:00,"[{'id': 227049, 'name': 'Kevin Galim', 'username': 'kev95', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/k/df788c/{size}.png', 'created_at': '2025-06-12T02:21:34.941Z', 'cooked': 'Hello,
\nI am trying to upload my recent arXiv paper (arXiv:2506.08373) to the HuggingFace Daily Papers platform, but I am encountering the following error:
\n{""error"":""Arxiv paper not found""}\nThe paper is publicly available on arXiv, so I’m not sure why it isn’t being recognized by the platform. Could you please help me resolve this issue?
\nThank you!
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-12T02:21:34.941Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 76, 'reads': 7, 'readers_count': 6, 'score': 386.4, 'yours': False, 'topic_id': 159000, 'topic_slug': 'unable-to-upload-arxiv-paper-to-huggingface-daily-papers', 'display_username': 'Kevin Galim', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://arxiv.org/abs/2506.08373', 'internal': False, 'reflection': False, 'title': '[2506.08373] Draft-based Approximate Inference for LLMs', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96744, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unable-to-upload-arxiv-paper-to-huggingface-daily-papers/159000/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 227053, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-12T02:48:41.745Z', 'cooked': 'I wonder if the Endpoint for submitting papers is malfunctioning… @pierric
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-12T02:48:41.745Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 6.4, 'yours': False, 'topic_id': 159000, 'topic_slug': 'unable-to-upload-arxiv-paper-to-huggingface-daily-papers', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/huggingface_hub/issues/2745', 'internal': False, 'reflection': False, 'title': '[HfApi] Add `submit_paper` endpoint · Issue #2745 · huggingface/huggingface_hub · GitHub', 'clicks': 8}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unable-to-upload-arxiv-paper-to-huggingface-daily-papers/159000/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 227209, 'name': 'Kevin Galim', 'username': 'kev95', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/k/df788c/{size}.png', 'created_at': '2025-06-13T02:07:09.420Z', 'cooked': 'It is working now. Thank you for your support!
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-13T02:07:09.420Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 6, 'readers_count': 5, 'score': 26.2, 'yours': False, 'topic_id': 159000, 'topic_slug': 'unable-to-upload-arxiv-paper-to-huggingface-daily-papers', 'display_username': 'Kevin Galim', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96744, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unable-to-upload-arxiv-paper-to-huggingface-daily-papers/159000/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 227281, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-13T14:08:06.126Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-13T14:08:06.126Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 159000, 'topic_slug': 'unable-to-upload-arxiv-paper-to-huggingface-daily-papers', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/unable-to-upload-arxiv-paper-to-huggingface-daily-papers/159000/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello,
+I am trying to upload my recent arXiv paper (arXiv:2506.08373) to the HuggingFace Daily Papers platform, but I am encountering the following error:
+{""error"":""Arxiv paper not found""}
+The paper is publicly available on arXiv, so I’m not sure why it isn’t being recognized by the platform. Could you please help me resolve this issue?
+Thank you!
",It is working now. Thank you for your support!
+Correct way to load multiple LoRA adapters for inference,https://discuss.huggingface.co/t/correct-way-to-load-multiple-lora-adapters-for-inference/158863,158863,9,2025-06-11 05:16:17.424000+00:00,"[{'id': 226879, 'name': 'Shruti Priya', 'username': 'sapphicart', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/90db22/{size}.png', 'created_at': '2025-06-11T05:16:17.482Z', 'cooked': 'I have trained two LoRA Adapters on top of the same base model. I saved the adapters with model.save_pretrained() Right now, I am trying to load both adapters for inference. My current approach is this:
base_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)\nmodel = PeftModelFromSequenceClassification.from_pretrained(base_model, adapter_1, adapter_name=""adapter_1"", num_labels=2)\n\nweighted_adapter_name=""two-lora""\nmodel.load_adapter(adapter_2, adapter_name=""adapter_2"")\n\nmodel.add_weighted_adapter(\n adapters=[""adapter_1"", ""adapter_2""],\n weights=[0.7, 0.3],\n adapter_name=weighted_adapter_name,\n combination_type=""linear"",\n)\nBut this gives me the error Cannot add weighted adapters if they target the same module with modules_to_save, but found 1 such instance(s).
Then, I tried another method from this documentation
\nbase_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)\nmodel = PeftMixedModel.from_pretrained(base_model, adapter_1, adapter_name=""adapter_1"")\n\nmodel.load_adapter(adapter_2, adapter_name=""adapter_2"")\nmodel.set_adapter([""adapter_1"", ""adapter_2""])\nBut this too throws an error ValueError: Only one adapter can be set at a time for modules_to_save.
I don’t understand what I am doing wrong. Should I try this:
\nget_peft_model with base_model and adapter_1add_adapter with adapter_2 to this modelBut with this approach how would I load both adapters for inference?
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-11T05:34:27.706Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 388, 'reads': 14, 'readers_count': 13, 'score': 1867.8, 'yours': False, 'topic_id': 158863, 'topic_slug': 'correct-way-to-load-multiple-lora-adapters-for-inference', 'display_username': 'Shruti Priya', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/peft/main/en/developer_guides/mixed_models', 'internal': False, 'reflection': False, 'title': 'Mixed adapter types', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95123, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/correct-way-to-load-multiple-lora-adapters-for-inference/158863/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 226880, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-11T05:35:43.348Z', 'cooked': 'Like this?
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-11T05:35:43.348Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 12, 'readers_count': 11, 'score': 32.4, 'yours': False, 'topic_id': 158863, 'topic_slug': 'correct-way-to-load-multiple-lora-adapters-for-inference', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/peft/discussions/1315', 'internal': False, 'reflection': False, 'title': 'How to train multiple LoRA adapters on the same base model concurrently. · huggingface/peft · Discussion #1315 · GitHub', 'clicks': 46}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/correct-way-to-load-multiple-lora-adapters-for-inference/158863/2', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226912, 'name': 'Shruti Priya', 'username': 'sapphicart', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/90db22/{size}.png', 'created_at': '2025-06-11T08:57:26.154Z', 'cooked': 'Thanks for the reply! I tried this and it works perfectly. But, when I try to save the model and load it from local directory, I get the error ValueError: Can\'t find \'adapter_config.json\' at \'/path/to/model\'. I have tried pushing the model to hub and then loading it, still the same error. I can see there is no adapter_config.json at the path. The json files are actually inside new directories for the adapters.
The file structure is like this:
\nmodel\n|____adapter_1\n| |_____adapter_config.json\n| |_____adapter_model.safetensors\n|____adapter_2\n| |_____adapter_config.json\n| |_____adapter_model.safetensors\n|____special_tokens_map.json\n|____tokenizer.json\n|____tokenizer.config.json\n|____vocab.txt\n|____README.md\nI am trying to load the model with adapters like this (the code is from this discussion):
\noutputs = ""/path/to/model""\nadapter_1 = ""/path/to/model/adapter_1""\nadapter_2 = ""/path/to/model/adapter_2""\n\nadapter_1_config = PeftConfig.from_pretrained(adapter_1)\nadapter_2_config = PeftConfig.from_pretrained(adapter_2)\n\nbase_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)\n\npeft_model = PeftModelForSequenceClassification.from_pretrained(base_model, outputs, num_labels=2)\npeft_model.load_adapter(adapter_1)\npeft_model.load_adapter(adapter_2)\nFound a solution!
\nInstead of loading PeftModel from base directory, I instead loaded it from adapter_1 then I loaded adapter_2 and used both for inference.
adapter_1 = ""/path/to/model/adapter_1""\nadapter_2 = ""/path/to/model/adapter_2""\n\nbase_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)\n\npeft_model = PeftModelForSequenceClassification.from_pretrained(base_model, adapter_1, num_labels=2)\npeft_model.load_adapter(adapter_1, adapter_name=""adapter_1"")\npeft_model.load_adapter(adapter_2, adapter_name=""adapter_2"")\npeft_model.base_model.set_adapter([""adapter_1"", ""adapter_2""])\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-11T21:20:26.083Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 1.4, 'yours': False, 'topic_id': 158863, 'topic_slug': 'correct-way-to-load-multiple-lora-adapters-for-inference', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/correct-way-to-load-multiple-lora-adapters-for-inference/158863/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have trained two LoRA Adapters on top of the same base model. I saved the adapters with model.save_pretrained() Right now, I am trying to load both adapters for inference. My current approach is this:
base_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)
+model = PeftModelFromSequenceClassification.from_pretrained(base_model, adapter_1, adapter_name=""adapter_1"", num_labels=2)
+
+weighted_adapter_name=""two-lora""
+model.load_adapter(adapter_2, adapter_name=""adapter_2"")
+
+model.add_weighted_adapter(
+ adapters=[""adapter_1"", ""adapter_2""],
+ weights=[0.7, 0.3],
+ adapter_name=weighted_adapter_name,
+ combination_type=""linear"",
+)
+But this gives me the error Cannot add weighted adapters if they target the same module with modules_to_save, but found 1 such instance(s).
Then, I tried another method from this documentation
+base_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)
+model = PeftMixedModel.from_pretrained(base_model, adapter_1, adapter_name=""adapter_1"")
+
+model.load_adapter(adapter_2, adapter_name=""adapter_2"")
+model.set_adapter([""adapter_1"", ""adapter_2""])
+But this too throws an error ValueError: Only one adapter can be set at a time for modules_to_save.
I don’t understand what I am doing wrong. Should I try this:
+get_peft_model with base_model and adapter_1add_adapter with adapter_2 to this modelBut with this approach how would I load both adapters for inference?
","Found a solution!
+Instead of loading PeftModel from base directory, I instead loaded it from adapter_1 then I loaded adapter_2 and used both for inference.
adapter_1 = ""/path/to/model/adapter_1""
+adapter_2 = ""/path/to/model/adapter_2""
+
+base_model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2, output_hidden_states=False)
+
+peft_model = PeftModelForSequenceClassification.from_pretrained(base_model, adapter_1, num_labels=2)
+peft_model.load_adapter(adapter_1, adapter_name=""adapter_1"")
+peft_model.load_adapter(adapter_2, adapter_name=""adapter_2"")
+peft_model.base_model.set_adapter([""adapter_1"", ""adapter_2""])
+Does anyone know about Linux? I’m trying to put a ISO on a flash drive
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T07:29:26.848Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 8, 'readers_count': 7, 'score': 26.6, 'yours': False, 'topic_id': 158545, 'topic_slug': 'linux-transfer-isos', 'display_username': 'Jordan kiss', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95898, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/linux-transfer-isos/158545/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 226431, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-09T08:03:07.654Z', 'cooked': 'I don’t know, but I found it when I searched.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T08:03:07.654Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 21.4, 'yours': False, 'topic_id': 158545, 'topic_slug': 'linux-transfer-isos', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://www.geeksforgeeks.org/techtips/setup-dual-boot-with-linux-and-windows/', 'internal': False, 'reflection': False, 'title': 'How to Set Up a Dual Boot with Ubuntu and Windows? - GeeksforGeeks', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/linux-transfer-isos/158545/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226536, 'name': 'Riley Fox', 'username': 'Mdrnfox', 'avatar_template': '/user_avatar/discuss.huggingface.co/mdrnfox/{size}/47695_2.png', 'created_at': '2025-06-09T17:53:17.498Z', 'cooked': 'Are you needing Linux? You could use a dual boot, VM, or download the WSL for windows.
\nI know you are going to need to burn the iso to the flash drive and format it with FAT32.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T17:53:17.498Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 31.0, 'yours': False, 'topic_id': 158545, 'topic_slug': 'linux-transfer-isos', 'display_username': 'Riley Fox', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94214, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/linux-transfer-isos/158545/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 226575, 'name': 'Jordan kiss', 'username': 'VexxaGlitch', 'avatar_template': '/user_avatar/discuss.huggingface.co/vexxaglitch/{size}/48728_2.png', 'created_at': '2025-06-09T21:22:12.199Z', 'cooked': 'I was trying to do it on a chrome book LOL but I was able to download it on a family members computer🫶🏼
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T21:22:12.199Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 158545, 'topic_slug': 'linux-transfer-isos', 'display_username': 'Jordan kiss', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95898, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/linux-transfer-isos/158545/4', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226701, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-10T09:22:17.178Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-10T09:22:17.178Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 158545, 'topic_slug': 'linux-transfer-isos', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/linux-transfer-isos/158545/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]",Does anyone know about Linux? I’m trying to put a ISO on a flash drive
,"Are you needing Linux? You could use a dual boot, VM, or download the WSL for windows.
+I know you are going to need to burn the iso to the flash drive and format it with FAT32.
" +How was self.loss_function implemented,https://discuss.huggingface.co/t/how-was-self-loss-function-implemented/158573,158573,9,2025-06-09 09:07:49.199000+00:00,"[{'id': 226460, 'name': 'Omar Samir', 'username': 'OmarSamir', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/c57346/{size}.png', 'created_at': '2025-06-09T09:07:49.255Z', 'cooked': 'Hi, I was curious about how the self.loss_function is implemented in the Qwen2.5-VL model to compute the loss during training.
\nCould someone explain how it works or point me to the relevant part of the code?
Here’s the link to the line I’m referring to:
\n\nThanks in advance!
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T09:07:49.255Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 235, 'reads': 11, 'readers_count': 10, 'score': 1117.0, 'yours': False, 'topic_id': 158573, 'topic_slug': 'how-was-self-loss-function-implemented', 'display_username': 'Omar Samir', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/blob/main/src/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py#L1615', 'internal': False, 'reflection': False, 'title': 'transformers/src/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py at main · huggingface/transformers · GitHub', 'clicks': 7}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96455, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-was-self-loss-function-implemented/158573/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 226478, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-09T11:13:52.136Z', 'cooked': 'Maybe this?
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T11:13:52.136Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 9, 'readers_count': 8, 'score': 56.6, 'yours': False, 'topic_id': 158573, 'topic_slug': 'how-was-self-loss-function-implemented', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/where-to-look-for-a-loss-definition-for-a-pretrained-model/26073', 'internal': True, 'reflection': False, 'title': 'Where to look for a loss definition for a pretrained model?', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/having-troubel-in-understanding-what-loss-is-currently-in-use/63395', 'internal': True, 'reflection': False, 'title': 'Having troubel in understanding what loss is currently in use', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-was-self-loss-function-implemented/158573/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226484, 'name': 'Omar Samir', 'username': 'OmarSamir', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/c57346/{size}.png', 'created_at': '2025-06-09T11:40:37.854Z', 'cooked': 'Thank you so much for sharing. However, these issues predated the Transformers version 4.53.0.dev0. What I want to know is where the self.loss_function was implemented for these models so I can modify it correctly.
\nThank you!
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-09T11:40:37.854Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 8, 'readers_count': 7, 'score': 46.4, 'yours': False, 'topic_id': 158573, 'topic_slug': 'how-was-self-loss-function-implemented', 'display_username': 'Omar Samir', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96455, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-was-self-loss-function-implemented/158573/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226495, 'name': 'Omar Samir', 'username': 'OmarSamir', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/o/c57346/{size}.png', 'created_at': '2025-06-09T12:32:19.186Z', 'cooked': 'The loss functions are defined in src/transformers/loss/loss_utils.py. The logic for selecting which loss function to use is implemented in the PreTrainedModel class, located in src/transformers/modeling_utils.py.
\nlink: transformers/src/transformers/loss/loss_utils.py at main · huggingface/transformers · GitHub
\nlink: transformers/src/transformers/modeling_utils.py at main · huggingface/transformers · GitHub
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-10T00:32:58.119Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 1.0, 'yours': False, 'topic_id': 158573, 'topic_slug': 'how-was-self-loss-function-implemented', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-was-self-loss-function-implemented/158573/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi, I was curious about how the self.loss_function is implemented in the Qwen2.5-VL model to compute the loss during training.
+Could someone explain how it works or point me to the relevant part of the code?
Here’s the link to the line I’m referring to:
+ +Thanks in advance!
","The loss functions are defined in src/transformers/loss/loss_utils.py. The logic for selecting which loss function to use is implemented in the PreTrainedModel class, located in src/transformers/modeling_utils.py.
+link: transformers/src/transformers/loss/loss_utils.py at main · huggingface/transformers · GitHub
+link: transformers/src/transformers/modeling_utils.py at main · huggingface/transformers · GitHub
I am a beginner with LLMs but I have been able to install Ollama, Oobabooga, sillytavern, anything llm, and convert between GGUF to GPTQ. I use windows 10 and Ubuntu 24.04 and also have some training experience with Flux on my home computer and Massed Compute.
\nI have been trying to train my own Lora using Oogbooga. I have tried on linux and windows. I have tried GGUF models and GPTQ models. I have tried .txt files and Json files generated from past chats. Nothing seems to work. I have also installed the Training Pro extension.
\nEvery time I try a GGUF model I receive the errpr:
\nAttribute Error: ‘LlamaServer’ object has no attribute ‘bos_token_id’
\nI was hoping that Training Pro would fix this error as it has a box to add a bos token to each data set item.
\nI get even more errors when trying to train a GPTQ model.
\nI have searched for alternate training.py files if that is the problem and have not found any that work.
\nI have not found much help on the internet or github.
\nAny suggestion?
\nThe whole console output for the Lora is:
\n16:24:07-798561 INFO Loaded “nvidia_Llama-3.1-Nemotron-Nano-4B-v1.1-Q6_K.gguf” in 2.51 seconds.
\n16:24:07-800568 INFO LOADER: “llama.cpp”
\n16:24:07-801571 INFO TRUNCATION LENGTH: 8192
\n16:24:07-802575 INFO INSTRUCTION TEMPLATE: “Custom (obtained from model metadata)”
\n16:24:23-882099 INFO Loading Text file…
\nPrecise raw text slicer: ON
\nSentences: 2967
\nText Blocks: 230
From a quick read of the code, I don’t think training a GGUF-quantized model is intended. How about trying it with the Transoformers-format model before GGUF quantization?
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-06T11:24:26.097Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 5, 'readers_count': 4, 'score': 36.0, 'yours': False, 'topic_id': 158175, 'topic_slug': 'unable-to-train-lora-with-oobabooga', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/oobabooga/text-generation-webui/blob/main/extensions/Training_PRO/script.py', 'internal': False, 'reflection': False, 'title': 'text-generation-webui/extensions/Training_PRO/script.py at main · oobabooga/text-generation-webui · GitHub', 'clicks': 7}, {'url': 'https://huggingface.co/nvidia/Llama-3.1-Nemotron-Nano-4B-v1.1', 'internal': False, 'reflection': False, 'title': 'nvidia/Llama-3.1-Nemotron-Nano-4B-v1.1 · Hugging Face', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unable-to-train-lora-with-oobabooga/158175/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226138, 'name': 'Chris', 'username': '363ls2gto', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/3/b3f665/{size}.png', 'created_at': '2025-06-07T03:24:50.274Z', 'cooked': 'Thank you for the reply. I also tried training using a transformers based GPTQ model. I received several errors attempting to train this format as well. I will try and get them posted. At least I know where not to waste my time now.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-07T03:24:50.274Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 5, 'readers_count': 4, 'score': 36.0, 'yours': False, 'topic_id': 158175, 'topic_slug': 'unable-to-train-lora-with-oobabooga', 'display_username': 'Chris', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96153, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unable-to-train-lora-with-oobabooga/158175/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226233, 'name': 'Chris', 'username': '363ls2gto', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/3/b3f665/{size}.png', 'created_at': '2025-06-07T21:49:28.446Z', 'cooked': 'I found the solution. I selected transformers but received errors. I was told to use pip-install XYZ (I can’t remember the exact command).
\nFor Ubuntu, run the cmd_linux.sh in konsole by right clicking and selecting this option. Make sure to select the “run in terminal” option vs “open terminal here” option. The cmd_linux.sh file is located in the same folder as the start.sh and update programs.
\nCopy the pip install command from oobabooga and paste it into the terminal you just opened. This command should be located in the bottom right portion of the page after all the previous errors listed in the training tab of the gradio.
\nYou have to do this a second time for a new package that also needs to be installed. This time oobabooga gives you an option of two different pip installs. Select the second option as the first does not work.
\nCopy and paste this new pip-install command that oobabooga gives you into the terminal. (you may have to close and restart the run in cmd_linux.sh terminal for the new pip install.)
\nIf you can load a GPTQ file using transformers, you should be able to train a LORA using either the normal or training pro extension.
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-06-07T21:54:27.020Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 5, 'readers_count': 4, 'score': 51.0, 'yours': False, 'topic_id': 158175, 'topic_slug': 'unable-to-train-lora-with-oobabooga', 'display_username': 'Chris', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96153, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/unable-to-train-lora-with-oobabooga/158175/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 226295, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-08T09:50:12.243Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-06-08T09:50:12.243Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 158175, 'topic_slug': 'unable-to-train-lora-with-oobabooga', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/unable-to-train-lora-with-oobabooga/158175/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am a beginner with LLMs but I have been able to install Ollama, Oobabooga, sillytavern, anything llm, and convert between GGUF to GPTQ. I use windows 10 and Ubuntu 24.04 and also have some training experience with Flux on my home computer and Massed Compute.
+I have been trying to train my own Lora using Oogbooga. I have tried on linux and windows. I have tried GGUF models and GPTQ models. I have tried .txt files and Json files generated from past chats. Nothing seems to work. I have also installed the Training Pro extension.
+Every time I try a GGUF model I receive the errpr:
+Attribute Error: ‘LlamaServer’ object has no attribute ‘bos_token_id’
+I was hoping that Training Pro would fix this error as it has a box to add a bos token to each data set item.
+I get even more errors when trying to train a GPTQ model.
+I have searched for alternate training.py files if that is the problem and have not found any that work.
+I have not found much help on the internet or github.
+Any suggestion?
+The whole console output for the Lora is:
+16:24:07-798561 INFO Loaded “nvidia_Llama-3.1-Nemotron-Nano-4B-v1.1-Q6_K.gguf” in 2.51 seconds.
+16:24:07-800568 INFO LOADER: “llama.cpp”
+16:24:07-801571 INFO TRUNCATION LENGTH: 8192
+16:24:07-802575 INFO INSTRUCTION TEMPLATE: “Custom (obtained from model metadata)”
+16:24:23-882099 INFO Loading Text file…
+Precise raw text slicer: ON
+Sentences: 2967
+Text Blocks: 230
I found the solution. I selected transformers but received errors. I was told to use pip-install XYZ (I can’t remember the exact command).
+For Ubuntu, run the cmd_linux.sh in konsole by right clicking and selecting this option. Make sure to select the “run in terminal” option vs “open terminal here” option. The cmd_linux.sh file is located in the same folder as the start.sh and update programs.
+Copy the pip install command from oobabooga and paste it into the terminal you just opened. This command should be located in the bottom right portion of the page after all the previous errors listed in the training tab of the gradio.
+You have to do this a second time for a new package that also needs to be installed. This time oobabooga gives you an option of two different pip installs. Select the second option as the first does not work.
+Copy and paste this new pip-install command that oobabooga gives you into the terminal. (you may have to close and restart the run in cmd_linux.sh terminal for the new pip install.)
+If you can load a GPTQ file using transformers, you should be able to train a LORA using either the normal or training pro extension.
" +Opus-MT: Translation returns <unk> token,https://discuss.huggingface.co/t/opus-mt-translation-returns-unk-token/158124,158124,13,2025-06-05 12:50:34.687000+00:00,"[{'id': 225882, 'name': 'Math Dons', 'username': 'mathdons', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/5e9695/{size}.png', 'created_at': '2025-06-05T12:50:34.757Z', 'cooked': '(x-posting with StackOverflow)
\nI’m having relatively good results with HelsinkiNlp models for translation, except for one thing: some special characters are omitted from the translation. If I decode without skipping the special tokens, I get the following:
\n<pad> <unk> a fait mal !</s>
<unk> is right where the translation should include a French Ç (expected result “Ça fait mal” from source “That hurts!”). Note:
<pad> APR<unk> S VOUS !</s> (should be “APRÈS VOUS !”)It’s definitely not a model issue, but a me issue, if I try on OpusTranslate Space (OPUS Translate - a Hugging Face Space by Helsinki-NLP), it works just fine.
\nI tried using the code verbatim from the model page, to no avail (Helsinki-NLP/opus-mt-tc-big-en-fr · Hugging Face)
\nMy current code is not far from it, and produces exactly the result I posted above:
\ndef __init__(self, model_path_or_name: str, source_language:str, target_langueg:str):\n self.device = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")\n self.tokenizer = MarianTokenizer.from_pretrained(model_path_or_name)\n self.model = MarianMTModel.from_pretrained(model_path_or_name).to(self.device)\n\ndef single_translate(self, text: str) -> str:\n """"""\n Translate a single sentence and return the translated string only.\n """"""\n inputs = self.tokenizer([text], return_tensors=""pt"", padding=True, truncation=True)\n input_ids = inputs.input_ids.to(self.model.device)\n with torch.no_grad():\n outputs = self.model.generate(input_ids=input_ids)\n decoded = self.tokenizer.batch_decode(outputs, skip_special_tokens=False)\n return decoded[0]\nAny advice would be greatly appreciated!
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-05T12:50:34.757Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 16, 'reads': 5, 'readers_count': 4, 'score': 96.0, 'yours': False, 'topic_id': 158124, 'topic_slug': 'opus-mt-translation-returns-unk-token', 'display_username': 'Math Dons', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/Helsinki-NLP/opus-translate', 'internal': False, 'reflection': False, 'title': 'OPUS Translate - a Hugging Face Space by Helsinki-NLP', 'clicks': 1}, {'url': 'https://huggingface.co/Helsinki-NLP/opus-mt-tc-big-en-fr', 'internal': False, 'reflection': False, 'title': 'Helsinki-NLP/opus-mt-tc-big-en-fr · Hugging Face', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96113, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/opus-mt-translation-returns-unk-token/158124/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 226047, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-06-06T12:58:25.566Z', 'cooked': 'It seems model issue…
\nfrom transformers import pipeline\npipe = pipeline(""translation"", model=""Helsinki-NLP/opus-mt-en-fr"")\nprint(pipe(""That hurts!"")) # [{\'translation_text\': \'Ça fait mal !\'}]\npipe = pipeline(""translation"", model=""Helsinki-NLP/opus-mt-tc-big-en-fr"")\nprint(pipe(""That hurts!"")) # [{\'translation_text\': \'a fait mal !\'}]\nDamn, it never occurred to me that the space could be using a different model in the same family/language. Thanks a lot, you’ve saved me a lot of headaches trying to find what was going wrong. Going to add a comment on the model / community page.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-06T13:37:55.045Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 21.0, 'yours': False, 'topic_id': 158124, 'topic_slug': 'opus-mt-translation-returns-unk-token', 'display_username': 'Math Dons', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96113, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/opus-mt-translation-returns-unk-token/158124/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 226132, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-07T01:38:40.309Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-07T01:38:40.309Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 158124, 'topic_slug': 'opus-mt-translation-returns-unk-token', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/opus-mt-translation-returns-unk-token/158124/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","(x-posting with StackOverflow)
+I’m having relatively good results with HelsinkiNlp models for translation, except for one thing: some special characters are omitted from the translation. If I decode without skipping the special tokens, I get the following:
+<pad> <unk> a fait mal !</s>
<unk> is right where the translation should include a French Ç (expected result “Ça fait mal” from source “That hurts!”). Note:
<pad> APR<unk> S VOUS !</s> (should be “APRÈS VOUS !”)It’s definitely not a model issue, but a me issue, if I try on OpusTranslate Space (OPUS Translate - a Hugging Face Space by Helsinki-NLP), it works just fine.
+I tried using the code verbatim from the model page, to no avail (Helsinki-NLP/opus-mt-tc-big-en-fr · Hugging Face)
+My current code is not far from it, and produces exactly the result I posted above:
+def __init__(self, model_path_or_name: str, source_language:str, target_langueg:str):
+ self.device = torch.device(""cuda"" if torch.cuda.is_available() else ""cpu"")
+ self.tokenizer = MarianTokenizer.from_pretrained(model_path_or_name)
+ self.model = MarianMTModel.from_pretrained(model_path_or_name).to(self.device)
+
+def single_translate(self, text: str) -> str:
+ """"""
+ Translate a single sentence and return the translated string only.
+ """"""
+ inputs = self.tokenizer([text], return_tensors=""pt"", padding=True, truncation=True)
+ input_ids = inputs.input_ids.to(self.model.device)
+ with torch.no_grad():
+ outputs = self.model.generate(input_ids=input_ids)
+ decoded = self.tokenizer.batch_decode(outputs, skip_special_tokens=False)
+ return decoded[0]
+Any advice would be greatly appreciated!
","It seems model issue…
+from transformers import pipeline
+pipe = pipeline(""translation"", model=""Helsinki-NLP/opus-mt-en-fr"")
+print(pipe(""That hurts!"")) # [{'translation_text': 'Ça fait mal !'}]
+pipe = pipeline(""translation"", model=""Helsinki-NLP/opus-mt-tc-big-en-fr"")
+print(pipe(""That hurts!"")) # [{'translation_text': 'a fait mal !'}]
+Hello Hugging Face community,
\nI’m planning to create a Hugging Face Space for real-time object detection, using Flask or FastAPI as the backend to process images or video streams with models like YOLO or DETR from the Hugging Face Space.
\nI have two questions:
\nIs it practical to run real-time object detection in a Space using Flask or FastAPI? What are the key limitations or best practices for deployment on Hugging Face Spaces?
\nI’m worried about violating Hugging Face’s policies. Could this type of Space risk my account being flagged or blocked? What steps can I take to ensure compliance with Hugging Face’s Terms of Service?
\nAny advice, example Spaces, or links to relevant documentation would be greatly appreciated. Thank you!
\nBest,
\nDanh Tran (danhtran2mind).
\n\n1
\n
I think Gradio’s backend is FastAPI, so I think it should be possible…
\nI don’t know much about Flask.
\n\n2
\n
I think 5. of this article mainly refers to prohibited acts in Spaces.
Hey, do you like cats. I love dogs. Thanks for your support.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-06-05T10:21:53.958Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 158020, 'topic_slug': 'can-i-build-a-real-time-object-detection-space-with-flask-or-fastapi-on-hugging-face', 'display_username': 'Danh Tran', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 96029, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-build-a-real-time-object-detection-space-with-flask-or-fastapi-on-hugging-face/158020/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 225953, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-05T22:22:49.286Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-06-05T22:22:49.286Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 2, 'readers_count': 1, 'score': 10.4, 'yours': False, 'topic_id': 158020, 'topic_slug': 'can-i-build-a-real-time-object-detection-space-with-flask-or-fastapi-on-hugging-face', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/can-i-build-a-real-time-object-detection-space-with-flask-or-fastapi-on-hugging-face/158020/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello Hugging Face community,
+I’m planning to create a Hugging Face Space for real-time object detection, using Flask or FastAPI as the backend to process images or video streams with models like YOLO or DETR from the Hugging Face Space.
+I have two questions:
+Is it practical to run real-time object detection in a Space using Flask or FastAPI? What are the key limitations or best practices for deployment on Hugging Face Spaces?
+I’m worried about violating Hugging Face’s policies. Could this type of Space risk my account being flagged or blocked? What steps can I take to ensure compliance with Hugging Face’s Terms of Service?
+Any advice, example Spaces, or links to relevant documentation would be greatly appreciated. Thank you!
+Best,
+Danh Tran (danhtran2mind).
++1
+
I think Gradio’s backend is FastAPI, so I think it should be possible…
+I don’t know much about Flask.
++2
+
I think 5. of this article mainly refers to prohibited acts in Spaces.
I need to distil whisper models. I have the python file that do that. It work in my pc, but i want to distil the large models.
\nI try to do that using the spaces (not free space) but i got the next message:
\nLaunch timed out space was not healthy after 30 min
\nHow to increment the launch time?
Maybe this setting?
\n\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-04T05:43:21.862Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 1.2, 'yours': False, 'topic_id': 157873, 'topic_slug': 'distil-whisper-models', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/hub/spaces-config-reference', 'internal': False, 'reflection': False, 'title': 'Spaces Configuration Reference', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/distil-whisper-models/157873/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 225694, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-04T17:43:51.330Z', 'cooked': '\n
startup_duration_timeout: string
\nSet a custom startup duration timeout for your Space. This is the maximum time your Space is allowed to start before it times out and is flagged as unhealthy. Defaults to 30 minutes, but any valid duration (like1h,30m) is acceptable.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-06-04T17:43:51.330Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 157873, 'topic_slug': 'distil-whisper-models', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/distil-whisper-models/157873/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I need to distil whisper models. I have the python file that do that. It work in my pc, but i want to distil the large models.
+I try to do that using the spaces (not free space) but i got the next message:
+Launch timed out space was not healthy after 30 min
+How to increment the launch time?
Maybe this setting?
+ + ++" +Adding labels from different files,https://discuss.huggingface.co/t/adding-labels-from-different-files/157864,157864,5,2025-06-03 16:34:10.583000+00:00,"[{'id': 225476, 'name': 'zacharia husain', 'username': 'zacharia-husain', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/z/90ced4/{size}.png', 'created_at': '2025-06-03T16:34:10.654Z', 'cooked': '+
startup_duration_timeout: string
+Set a custom startup duration timeout for your Space. This is the maximum time your Space is allowed to start before it times out and is flagged as unhealthy. Defaults to 30 minutes, but any valid duration (like1h,30m) is acceptable.
If I have multiple texts in a folder and a csv file with token classification labels, how would I merge them together so when I index the dataset the text and labels will be in the same index (like how in the examples the imdb dataset has sentiment and text at the same index). My understanding is that you can only pass one file type to load_datasets, and map I cant figure out how to use map when the size of the labels varies (it depends on amount of tokens).
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-06-03T16:34:10.654Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 6, 'readers_count': 5, 'score': 66.2, 'yours': False, 'topic_id': 157864, 'topic_slug': 'adding-labels-from-different-files', 'display_username': 'zacharia husain', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95904, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/adding-labels-from-different-files/157864/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 225479, 'name': 'Riley Fox', 'username': 'Mdrnfox', 'avatar_template': '/user_avatar/discuss.huggingface.co/mdrnfox/{size}/47695_2.png', 'created_at': '2025-06-03T16:48:56.739Z', 'cooked': '\nWhat I would do is:
\nRead in your files
\nAlign your labels to your tokenized text. Try using tokenizer(…, return_offsets_mapping=True) helps you align labels to tokens.
\nThen create a dataset object manually.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-06-04T14:58:44.199Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 157864, 'topic_slug': 'adding-labels-from-different-files', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/adding-labels-from-different-files/157864/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","If I have multiple texts in a folder and a csv file with token classification labels, how would I merge them together so when I index the dataset the text and labels will be in the same index (like how in the examples the imdb dataset has sentiment and text at the same index). My understanding is that you can only pass one file type to load_datasets, and map I cant figure out how to use map when the size of the labels varies (it depends on amount of tokens).
"," +What I would do is:
+Read in your files
+Align your labels to your tokenized text. Try using tokenizer(…, return_offsets_mapping=True) helps you align labels to tokens.
+Then create a dataset object manually.
Hello!
I’m benchmarking inference performance using Whisper and the .generate() method, switching between using/not using the k-v cache).
My understanding is that when using the cache, inference should be faster (since we don’t recompute k-v states and cache them instead), but VRAM usage higher (since we keep the cached tensors in memory).
\nHowever, I’m finding that when using cache that inference is faster, but VRAM stays the same
Here are my results with/without cache for the tiny and base Whisper checkpoints:
\n| \n | Inf time with | \nInf time without | \nVRAM with | \nVRAM without | \n
|---|---|---|---|---|
| tiny | \n9.0 | \n12.0 | \n1381 | \n1381 | \n
| base | \n11.3 | \n18.4 | \n1523 | \n1523 | \n
These experiments are run with greedy decoding, batch size of 1 and 73 eval samples on a 16GB V100. I’m computing VRAM by calling nvidia-smi and monitoring how much usage there is on the GPU.
Is this as expected? Or should we see lower VRAM without cache?
\nNotebook: codesnippets/benchmark_whisper_cache.ipynb at main · sanchit-gandhi/codesnippets · GitHub
\nfrom datasets import load_dataset\nfrom transformers import WhisperConfig, WhisperForConditionalGeneration, WhisperProcessor\n\nimport torch\nfrom torch.utils.data import DataLoader\nimport numpy as np\n\nimport time\nfrom tqdm import tqdm\nimport subprocess as sp\nimport os\nimport sched\n\ncheckpoint_id = ""openai/whisper-tiny.en""\nprocessor = WhisperProcessor.from_pretrained(checkpoint_id)\n\nmodel = WhisperForConditionalGeneration.from_pretrained(checkpoint_id)\nmodel.to(""cuda"")\nmodel.half()\n\nlibrispeech = load_dataset(""hf-internal-testing/librispeech_asr_dummy"", ""clean"", split=""validation"")\n\ndef preprocess(batch): \n batch[""input_features""] = processor(batch[""audio""][""array""], sampling_rate=16000, return_tensors=""pt"").input_features[0]\n return batch\n\ndataset_processed = librispeech.map(preprocess, remove_columns=librispeech.column_names)\n\ndataloader = DataLoader(dataset_processed.with_format(""torch""), batch_size=1)\n\n\ndef get_gpu_memory():\n """"""\n Python equivalent of nvidia-smi, copied from https://stackoverflow.com/a/67722676\n and verified as being equivalent ✅\n """"""\n output_to_list = lambda x: x.decode(\'ascii\').split(\'\\n\')[:-1]\n \n COMMAND = ""nvidia-smi --query-gpu=memory.used --format=csv""\n \n try:\n memory_use_info = output_to_list(sp.check_output(COMMAND.split(),stderr=sp.STDOUT))[1:]\n \n except sp.CalledProcessError as e:\n raise RuntimeError(""command \'{}\' return with error (code {}): {}"".format(e.cmd, e.returncode, e.output))\n \n memory_use_values = [int(x.split()[0]) for i, x in enumerate(memory_use_info)]\n return memory_use_values\n\n# benchmark generation with cache\n\nstart = time.time()\nfor batch in tqdm(dataloader):\n predicted_ids = model.generate(batch[""input_features""].to(""cuda"").half(), max_new_tokens=128, use_cache=True)\nruntime = time.time() - start\n\nprint(""Runtime with: "", runtime)\nprint(""VRAM with: "", get_gpu_memory()[0])\n\n# if we don\'t delete and re-load the model the GPU use is lower the second time round: warm-up effects?\ndel model\ntorch.cuda.empty_cache()\n\n# benchmark without cache\n\nmodel = WhisperForConditionalGeneration.from_pretrained(checkpoint_id)\nmodel.to(""cuda"")\nmodel.half()\n\nstart = time.time()\nfor batch in tqdm(dataloader):\n predicted_ids = model.generate(batch[""input_features""].to(""cuda"").half(), max_new_tokens=128, use_cache=False)\nruntime = time.time() - start\n\nprint(""Runtime without: "", runtime)\nprint(""VRAM without: "", get_gpu_memory()[0])\nPrint Output:
\nRuntime with: 8.990428924560547\nVRAM with: 1381\nRuntime without: 11.993675231933594\nVRAM without: 1381\nThanks!
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2023-02-08T10:05:24.408Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 15561, 'reads': 249, 'readers_count': 248, 'score': 77799.8, 'yours': False, 'topic_id': 31272, 'topic_slug': 'generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage', 'display_username': 'Sanchit Gandhi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 6, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/main_classes/text_generation#transformers.GenerationConfig.use_cache', 'internal': False, 'reflection': False, 'title': 'Generation', 'clicks': 1346}, {'url': 'https://github.com/sanchit-gandhi/codesnippets/blob/main/benchmark_whisper_cache.ipynb', 'internal': False, 'reflection': False, 'title': 'codesnippets/benchmark_whisper_cache.ipynb at main · sanchit-gandhi/codesnippets · GitHub', 'clicks': 297}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 9227, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage/31272/1', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 57335, 'name': 'Patrick von Platen', 'username': 'patrickvonplaten', 'avatar_template': '/user_avatar/discuss.huggingface.co/patrickvonplaten/{size}/2171_2.png', 'created_at': '2023-02-08T11:56:56.097Z', 'cooked': 'Nice write-up!
\nI think the decoder sequence length and the hidden states of the model might be too small to see a difference here in VRAM.
\nThe reason VRAM should be higher when caching the k,v states is because we cache the projected k,v states of every layer. This means that our cache is of size:
\n2 * (hidden_size) * (num_layers) * (decoder_length)
\nFor VRAM computation, this memory is more or less always added to the peak memory of the computation graph.
\nFor comparison, we don’t have this memory when not caching. The memory we always have when not caching before doing the attention QK^T computation (which is probs the bottleneck) is 2 * (hidden_size) * 1 * (decoder_length) . Those are the q, v states right that are computed during attention.
\n=> I expect that here (num_layers), (hidden_size) and (decoder_length) are too small to make a difference.
\nThe easiest thing to check here would be to use a bigger model and generate to much longer (set eos to None and generate to 256 tokens).
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2023-02-08T11:56:56.097Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 238, 'reads': 204, 'readers_count': 203, 'score': 1260.8, 'yours': False, 'topic_id': 31272, 'topic_slug': 'generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage', 'display_username': 'Patrick von Platen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 170, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage/31272/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 57336, 'name': 'Patrick von Platen', 'username': 'patrickvonplaten', 'avatar_template': '/user_avatar/discuss.huggingface.co/patrickvonplaten/{size}/2171_2.png', 'created_at': '2023-02-08T11:58:02.142Z', 'cooked': 'Overall this is an interesting finding though as it means that the k,v cache probably doesn’t play a big role in reducing VRAM for ASR and at that model size.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2023-02-08T11:58:02.142Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 43, 'reads': 187, 'readers_count': 186, 'score': 252.4, 'yours': False, 'topic_id': 31272, 'topic_slug': 'generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage', 'display_username': 'Patrick von Platen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 170, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage/31272/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 57349, 'name': 'Joao Gante', 'username': 'joaogante', 'avatar_template': '/user_avatar/discuss.huggingface.co/joaogante/{size}/20106_2.png', 'created_at': '2023-02-08T13:29:29.546Z', 'cooked': '@sanchit-gandhi a few extra numbers – modifying your script to run on GPT-J with FP16 on an 3090, with input_ids.shape[1]=16 and max_new_tokens=256, we get:
14071MB of GPU usage with use_cache=False\n13233MB of GPU usage with use_cache=True\nThe difference becomes more visible with large models and large sequence lengths
Thank you very much for the detailed response!
\nThat makes sense that the difference in VRAM with/without using cache is not significant for a model with such low dimensionality.
\nRepeating the experiment with the large-v2 checkpoint (hidden_size=1280, num_layers=32) and generating to 256 tokens yields measurable differences in VRAM, albeit still only marginal:
\nVRAM with: 7597\nVRAM without: 7515\nDiff: 82\n(all values in MB)
\nAs we expect, the effect is amplified at 512 tokens, scaling (almost) linearly with decoder_length:
VRAM with: 7639\nVRAM without: 7519\nDiff: 120\nASR models tend to generate quite short decoder-lengths. For example, the average token length in the LibriSpeech validation corpus is just ~20 tokens. Setting the max length accordingly, we get:
\nVRAM with: 7515\nVRAM without: 7511\nDiff: 4\nSo pretty insignificant! My intuition is that since VRAM difference with/without cache is proportional to decoder-length, k-v cache doesn’t have a big effect on VRAM for ASR models, even for larger checkpoints.
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2023-02-08T14:21:33.999Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 220, 'reads': 164, 'readers_count': 163, 'score': 1112.8, 'yours': False, 'topic_id': 31272, 'topic_slug': 'generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage', 'display_username': 'Sanchit Gandhi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 9227, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage/31272/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 225509, 'name': 'vhr', 'username': 'vhr1007', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/v/8e8cbc/{size}.png', 'created_at': '2025-06-03T21:25:14.414Z', 'cooked': 'Good Analysis, but generally you need to monitor max_cuda_allocation to know the max memory choke point in inference call, that will know usage of VRAM,
', 'post_number': 6, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-06-03T21:25:14.414Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 20.6, 'yours': False, 'topic_id': 31272, 'topic_slug': 'generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage', 'display_username': 'vhr', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95926, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/generate-using-k-v-cache-is-faster-but-no-difference-to-memory-usage/31272/6', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello!
I’m benchmarking inference performance using Whisper and the .generate() method, switching between using/not using the k-v cache).
My understanding is that when using the cache, inference should be faster (since we don’t recompute k-v states and cache them instead), but VRAM usage higher (since we keep the cached tensors in memory).
+However, I’m finding that when using cache that inference is faster, but VRAM stays the same
Here are my results with/without cache for the tiny and base Whisper checkpoints:
+| + | Inf time with | +Inf time without | +VRAM with | +VRAM without | +
|---|---|---|---|---|
| tiny | +9.0 | +12.0 | +1381 | +1381 | +
| base | +11.3 | +18.4 | +1523 | +1523 | +
These experiments are run with greedy decoding, batch size of 1 and 73 eval samples on a 16GB V100. I’m computing VRAM by calling nvidia-smi and monitoring how much usage there is on the GPU.
Is this as expected? Or should we see lower VRAM without cache?
+Notebook: codesnippets/benchmark_whisper_cache.ipynb at main · sanchit-gandhi/codesnippets · GitHub
+from datasets import load_dataset
+from transformers import WhisperConfig, WhisperForConditionalGeneration, WhisperProcessor
+
+import torch
+from torch.utils.data import DataLoader
+import numpy as np
+
+import time
+from tqdm import tqdm
+import subprocess as sp
+import os
+import sched
+
+checkpoint_id = ""openai/whisper-tiny.en""
+processor = WhisperProcessor.from_pretrained(checkpoint_id)
+
+model = WhisperForConditionalGeneration.from_pretrained(checkpoint_id)
+model.to(""cuda"")
+model.half()
+
+librispeech = load_dataset(""hf-internal-testing/librispeech_asr_dummy"", ""clean"", split=""validation"")
+
+def preprocess(batch):
+ batch[""input_features""] = processor(batch[""audio""][""array""], sampling_rate=16000, return_tensors=""pt"").input_features[0]
+ return batch
+
+dataset_processed = librispeech.map(preprocess, remove_columns=librispeech.column_names)
+
+dataloader = DataLoader(dataset_processed.with_format(""torch""), batch_size=1)
+
+
+def get_gpu_memory():
+ """"""
+ Python equivalent of nvidia-smi, copied from https://stackoverflow.com/a/67722676
+ and verified as being equivalent ✅
+ """"""
+ output_to_list = lambda x: x.decode('ascii').split('\n')[:-1]
+
+ COMMAND = ""nvidia-smi --query-gpu=memory.used --format=csv""
+
+ try:
+ memory_use_info = output_to_list(sp.check_output(COMMAND.split(),stderr=sp.STDOUT))[1:]
+
+ except sp.CalledProcessError as e:
+ raise RuntimeError(""command '{}' return with error (code {}): {}"".format(e.cmd, e.returncode, e.output))
+
+ memory_use_values = [int(x.split()[0]) for i, x in enumerate(memory_use_info)]
+ return memory_use_values
+
+# benchmark generation with cache
+
+start = time.time()
+for batch in tqdm(dataloader):
+ predicted_ids = model.generate(batch[""input_features""].to(""cuda"").half(), max_new_tokens=128, use_cache=True)
+runtime = time.time() - start
+
+print(""Runtime with: "", runtime)
+print(""VRAM with: "", get_gpu_memory()[0])
+
+# if we don't delete and re-load the model the GPU use is lower the second time round: warm-up effects?
+del model
+torch.cuda.empty_cache()
+
+# benchmark without cache
+
+model = WhisperForConditionalGeneration.from_pretrained(checkpoint_id)
+model.to(""cuda"")
+model.half()
+
+start = time.time()
+for batch in tqdm(dataloader):
+ predicted_ids = model.generate(batch[""input_features""].to(""cuda"").half(), max_new_tokens=128, use_cache=False)
+runtime = time.time() - start
+
+print(""Runtime without: "", runtime)
+print(""VRAM without: "", get_gpu_memory()[0])
+Print Output:
+Runtime with: 8.990428924560547
+VRAM with: 1381
+Runtime without: 11.993675231933594
+VRAM without: 1381
+Thanks!
","Nice write-up!
+I think the decoder sequence length and the hidden states of the model might be too small to see a difference here in VRAM.
+The reason VRAM should be higher when caching the k,v states is because we cache the projected k,v states of every layer. This means that our cache is of size:
+2 * (hidden_size) * (num_layers) * (decoder_length)
+For VRAM computation, this memory is more or less always added to the peak memory of the computation graph.
+For comparison, we don’t have this memory when not caching. The memory we always have when not caching before doing the attention QK^T computation (which is probs the bottleneck) is 2 * (hidden_size) * 1 * (decoder_length) . Those are the q, v states right that are computed during attention.
+=> I expect that here (num_layers), (hidden_size) and (decoder_length) are too small to make a difference.
+The easiest thing to check here would be to use a bigger model and generate to much longer (set eos to None and generate to 256 tokens).
" +What are the most effective recent approaches for predicting social media post virality?,https://discuss.huggingface.co/t/what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality/157384,157384,13,2025-05-30 13:30:44.236000+00:00,"[{'id': 224822, 'name': 'DB', 'username': 'catpawws', 'avatar_template': '/user_avatar/discuss.huggingface.co/catpawws/{size}/48526_2.png', 'created_at': '2025-05-30T13:30:44.300Z', 'cooked': 'I’m currently working on a project related to virality prediction . I came across this 2024 paper that combines BERT and CNN for Twitter virality classification:
\n Virality Prediction on Twitter Using Combined CNN and BERT Models | IEEE Xplore
Do you think this BERT+CNN hybrid is a good choice in 2024/2025?
\nAre there more advanced or better-performing models (e.g. graph-based, transformer-only, multimodal) that you’d recommend for this task?
Any suggestions or insights from your experience would be greatly appreciated!
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-30T13:30:44.300Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 52, 'reads': 7, 'readers_count': 6, 'score': 271.4, 'yours': False, 'topic_id': 157384, 'topic_slug': 'what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality', 'display_username': 'DB', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://ieeexplore.ieee.org/document/10913355', 'internal': False, 'reflection': False, 'title': 'Virality Prediction on Twitter Using Combined CNN and BERT Models | IEEE Conference Publication | IEEE Xplore', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95548, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality/157384/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 224888, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-30T23:48:53.073Z', 'cooked': 'I can’t find any methods other than BERT-based models…
\nhttps://www.researchgate.net/publication/355473219_Virality_Prediction_for_News_Tweets_Using_RoBERTa
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-30T23:48:53.073Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 5, 'readers_count': 4, 'score': 31.0, 'yours': False, 'topic_id': 157384, 'topic_slug': 'what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://ar5iv.labs.arxiv.org/html/2303.06120', 'internal': False, 'reflection': False, 'title': '[2303.06120] Measuring and Detecting Virality on Social Media: The Case of Twitter’s Viral Tweets Topic', 'clicks': 2}, {'url': 'https://www.researchgate.net/publication/355473219_Virality_Prediction_for_News_Tweets_Using_RoBERTa', 'internal': False, 'reflection': False, 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality/157384/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 225182, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-06-02T09:44:35.310Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-06-02T09:44:35.310Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 157384, 'topic_slug': 'what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/what-are-the-most-effective-recent-approaches-for-predicting-social-media-post-virality/157384/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m currently working on a project related to virality prediction . I came across this 2024 paper that combines BERT and CNN for Twitter virality classification:
+ Virality Prediction on Twitter Using Combined CNN and BERT Models | IEEE Xplore
Do you think this BERT+CNN hybrid is a good choice in 2024/2025?
+Are there more advanced or better-performing models (e.g. graph-based, transformer-only, multimodal) that you’d recommend for this task?
Any suggestions or insights from your experience would be greatly appreciated!
","I can’t find any methods other than BERT-based models…
+https://www.researchgate.net/publication/355473219_Virality_Prediction_for_News_Tweets_Using_RoBERTa
" +AI Agent Course,https://discuss.huggingface.co/t/ai-agent-course/157406,157406,21,2025-05-30 16:10:43.005000+00:00,"[{'id': 224848, 'name': 'Chan Kam Wing', 'username': 'WingNeville', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/w/e9a140/{size}.png', 'created_at': '2025-05-30T16:10:43.082Z', 'cooked': 'Hi everyone,
\nI’m currently running this notebook:
\nunit2/smolagents/code_agents.ipynb · agents-course/notebooks at main, but it’s returning an error.
So far, I’ve been unable to successfully run most of the examples in the course. I’m unsure if this is due to an issue with my account settings.
\nDo you have any suggestions?
\nValueError Traceback (most recent call last)
\n/usr/local/lib/python3.11/dist-packages/smolagents/agents.py in _step_stream(self, memory_step)
\n1495 else:
\n → 1496 chat_message: ChatMessage = self.model.generate(
\n1497 input_messages,
8 frames
\nValueError: Provider ‘nscale’ not supported. Available values: ‘auto’ or any provider from [‘black-forest-labs’, ‘cerebras’, ‘cohere’, ‘fal-ai’, ‘fireworks-ai’, ‘hf-inference’, ‘hyperbolic’, ‘nebius’, ‘novita’, ‘openai’, ‘replicate’, ‘sambanova’, ‘together’].Passing ‘auto’ (default value) will automatically select the first provider available for the model, sorted by the user’s order in Hugging Face – The AI community building the future..
The above exception was the direct cause of the following exception:
\nAgentGenerationError Traceback (most recent call last)
\n/usr/local/lib/python3.11/dist-packages/smolagents/agents.py in _step_stream(self, memory_step)
\n1516 memory_step.model_output = output_text
\n1517 except Exception as e:
\n → 1518 raise AgentGenerationError(f""Error in generating model output:\\n{e}"", self.logger) from e
\n1519
\n1520 ### Parse output ###
AgentGenerationError: Error in generating model output:
\nProvider ‘nscale’ not supported. Available values: ‘auto’ or any provider from [‘black-forest-labs’, ‘cerebras’, ‘cohere’, ‘fal-ai’, ‘fireworks-ai’, ‘hf-inference’, ‘hyperbolic’, ‘nebius’, ‘novita’, ‘openai’, ‘replicate’, ‘sambanova’, ‘together’].Passing ‘auto’ (default value) will automatically select the first provider available for the model, sorted by the user’s order in Hugging Face – The AI community building the future..
You are trying to use a provider called NScale. The backend doesn’t support that provider for that Model. Switch to auto and Huggingface will pick the first provider for you for that model.
\nAlternatively, you can go research the model on Huggingface and see what providers are available for that model and pass that arguement accordingly.
Hope that helps
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-31T06:41:50.658Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 20, 'readers_count': 19, 'score': 4.0, 'yours': False, 'topic_id': 157406, 'topic_slug': 'ai-agent-course', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/ai-agent-course/157406/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I’m currently running this notebook:
+unit2/smolagents/code_agents.ipynb · agents-course/notebooks at main, but it’s returning an error.
So far, I’ve been unable to successfully run most of the examples in the course. I’m unsure if this is due to an issue with my account settings.
+Do you have any suggestions?
+ValueError Traceback (most recent call last)
+/usr/local/lib/python3.11/dist-packages/smolagents/agents.py in _step_stream(self, memory_step)
+1495 else:
+ → 1496 chat_message: ChatMessage = self.model.generate(
+1497 input_messages,
8 frames
+ValueError: Provider ‘nscale’ not supported. Available values: ‘auto’ or any provider from [‘black-forest-labs’, ‘cerebras’, ‘cohere’, ‘fal-ai’, ‘fireworks-ai’, ‘hf-inference’, ‘hyperbolic’, ‘nebius’, ‘novita’, ‘openai’, ‘replicate’, ‘sambanova’, ‘together’].Passing ‘auto’ (default value) will automatically select the first provider available for the model, sorted by the user’s order in Hugging Face – The AI community building the future..
The above exception was the direct cause of the following exception:
+AgentGenerationError Traceback (most recent call last)
+/usr/local/lib/python3.11/dist-packages/smolagents/agents.py in _step_stream(self, memory_step)
+1516 memory_step.model_output = output_text
+1517 except Exception as e:
+ → 1518 raise AgentGenerationError(f""Error in generating model output:\n{e}"", self.logger) from e
+1519
+1520 ### Parse output ###
AgentGenerationError: Error in generating model output:
+Provider ‘nscale’ not supported. Available values: ‘auto’ or any provider from [‘black-forest-labs’, ‘cerebras’, ‘cohere’, ‘fal-ai’, ‘fireworks-ai’, ‘hf-inference’, ‘hyperbolic’, ‘nebius’, ‘novita’, ‘openai’, ‘replicate’, ‘sambanova’, ‘together’].Passing ‘auto’ (default value) will automatically select the first provider available for the model, sorted by the user’s order in Hugging Face – The AI community building the future..
You are trying to use a provider called NScale. The backend doesn’t support that provider for that Model. Switch to auto and Huggingface will pick the first provider for you for that model.
+Alternatively, you can go research the model on Huggingface and see what providers are available for that model and pass that arguement accordingly.
Hope that helps
Here’s the error I’m seeing for Container logs:
\nError: Failed to load logs: Not Found. Logs are persisted for 30 days after the Space stops running.
', 'post_number': 1, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T18:13:54.291Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2222, 'reads': 105, 'readers_count': 104, 'score': 10721.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Dan Moen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/spaces-and-building-stuck-infra-side-issue-and-how-to-troubleshoot-further/54158/5', 'internal': True, 'reflection': True, 'title': 'Spaces and ""Building"" stuck, infra side issue and how to troubleshoot further?', 'clicks': 3}, {'url': 'https://discuss.huggingface.co/t/error-failed-to-load-logs-not-found-logs-are-persisted-for-30-days-after-the-space-stops-running/66922/4', 'internal': True, 'reflection': True, 'title': 'Error: Failed to load logs: Not Found. Logs are persisted for 30 days after the Space stops running', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28476, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/1', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 88645, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-09-08T18:24:27.043Z', 'cooked': 'hi @155elkhorn could you please share more details? do you have a public Space link to share? thanks
', 'post_number': 2, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T18:24:27.043Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 19, 'reads': 101, 'readers_count': 100, 'score': 110.2, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 88668, 'name': 'Dan Moen', 'username': '155elkhorn', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png', 'created_at': '2023-09-08T22:51:21.783Z', 'cooked': 'The space isn’t public, but here’s the link to the space: https://huggingface.co/spaces/PikeAndVine/SD-Inpaint-POC
', 'post_number': 3, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T22:51:21.783Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 95, 'readers_count': 94, 'score': 39.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Dan Moen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/PikeAndVine/SD-Inpaint-POC', 'internal': False, 'reflection': False, 'clicks': 98}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28476, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 88669, 'name': 'Dan Moen', 'username': '155elkhorn', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png', 'created_at': '2023-09-08T22:52:19.507Z', 'cooked': 'I went ahead and made it public for now in case that helps.
', 'post_number': 4, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T22:52:19.507Z', 'reply_count': 1, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 94, 'readers_count': 93, 'score': 48.8, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Dan Moen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 28476, 'username': '155elkhorn', 'name': 'Dan Moen', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28476, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 88670, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-09-08T23:04:09.045Z', 'cooked': 'thanks for sharing, I duplicate your Space for testing purposes and it build and run normally
\nCould you please try a Factory Reboot?
\nAnother tip is, if you’re using the persistent data you set set HF_HOME to /data/.huggingface So you won’t need to re-download models every new build
I’ve done at least 5 factory reboots. I tried another one and here’s the error I’m getting:
\nBuild logs:
\n===== Build Queued at 2023-09-08 23:07:41 / Commit SHA: fd2693c =====\n\n--> FROM docker.io/nvidia/cuda:11.3.1-cudnn8-devel-ubuntu18.04@sha256:69cd988555eabe116f76acc754b363eee75f37674c23adb2b523f5fa32543984\nDONE 29.1s\n\n--> RUN apt-get update && apt-get install -y git make build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm libncursesw5-dev xz-utils tk-dev libxml2-dev libxmlsec1-dev libffi-dev liblzma-dev git-lfs \tffmpeg libsm6 libxext6 cmake libgl1-mesa-glx \t\t&& rm -rf /var/lib/apt/lists/* \t&& git lfs install\n\n--> ERROR: failed commit on ref ""layer-sha256:c89166c8ea49f8e433445b622e665a321cf96442e5a4b86ca0d3d2b2812a8b6d"": unexpected commit digest sha256:0f494b781dd9bb64e7fff4a96d5be6526ca5b57377c14a5c2c572edbc3d8f6a4, expected sha256:c89166c8ea49f8e433445b622e665a321cf96442e5a4b86ca0d3d2b2812a8b6d: failed precondition\nSorry, that’s very odd. Did you just duplicated it and got that error? Are you using persistent storage?
', 'post_number': 7, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T23:12:31.403Z', 'reply_count': 1, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 70, 'readers_count': 69, 'score': 24.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 28476, 'username': '155elkhorn', 'name': 'Dan Moen', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 88678, 'name': 'Dan Moen', 'username': '155elkhorn', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png', 'created_at': '2023-09-08T23:18:51.265Z', 'cooked': 'I just made a copy like you did and it actually started, yay!
\nYes, I have persistent storage turned on and I added that HF_HOME variable like you suggested.
', 'post_number': 8, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T23:18:51.265Z', 'reply_count': 1, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 72, 'readers_count': 71, 'score': 64.4, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Dan Moen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28476, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/8', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 88680, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-09-08T23:19:54.357Z', 'cooked': 'Sorry, for the issues, next week we could have @chris-rannou to have a look on the infra side thanks
', 'post_number': 9, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T23:19:54.357Z', 'reply_count': 0, 'reply_to_post_number': 8, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 72, 'readers_count': 71, 'score': 34.4, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 28476, 'username': '155elkhorn', 'name': 'Dan Moen', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 88681, 'name': 'Dan Moen', 'username': '155elkhorn', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png', 'created_at': '2023-09-08T23:20:28.714Z', 'cooked': 'I have quite a few scripts pointed at this space via API, so would really prefer to get it running versus moving over to the copy.
', 'post_number': 10, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-09-08T23:20:28.714Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 70, 'readers_count': 69, 'score': 94.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Dan Moen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28476, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/10', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 94166, 'name': 'George', 'username': 'wholewhale', 'avatar_template': '/user_avatar/discuss.huggingface.co/wholewhale/{size}/20295_2.png', 'created_at': '2023-10-12T21:13:19.761Z', 'cooked': 'I am getting the same Log error and build failure. Chat with PDF • OpenAI - a Hugging Face Space by wholewhale
', 'post_number': 11, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-12T21:13:19.761Z', 'reply_count': 1, 'reply_to_post_number': 10, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 61, 'readers_count': 60, 'score': 42.2, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'George', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/wholewhale/causewriter-chat-with-pdf-openai?logs=build', 'internal': False, 'reflection': False, 'title': 'Chat with PDF •\xa0OpenAI - a Hugging Face Space by wholewhale', 'clicks': 15}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 28476, 'username': '155elkhorn', 'name': 'Dan Moen', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31052, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/11', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 94169, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-10-12T21:30:15.099Z', 'cooked': 'Apologies, we had some internal issues on our infra, could you please try rebooting/factory rebooting now?
', 'post_number': 12, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-12T21:30:15.099Z', 'reply_count': 1, 'reply_to_post_number': 11, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 60, 'readers_count': 59, 'score': 27.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 31052, 'username': 'wholewhale', 'name': 'George', 'avatar_template': '/user_avatar/discuss.huggingface.co/wholewhale/{size}/20295_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/12', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 94170, 'name': 'George', 'username': 'wholewhale', 'avatar_template': '/user_avatar/discuss.huggingface.co/wholewhale/{size}/20295_2.png', 'created_at': '2023-10-12T21:32:10.662Z', 'cooked': 'Getting: "" 500
\nInternal Error - We’re working hard to fix this as soon as possible!""
\n(TY for the quick reply)
', 'post_number': 13, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-12T21:32:10.662Z', 'reply_count': 1, 'reply_to_post_number': 12, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 63, 'readers_count': 62, 'score': 37.6, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'George', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31052, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/13', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 94171, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-10-12T21:39:44.083Z', 'cooked': '\nApologies, we’re in recovery mode, I’ll ping when things are back
', 'post_number': 14, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-12T21:39:44.083Z', 'reply_count': 2, 'reply_to_post_number': 13, 'quote_count': 1, 'incoming_link_count': 1, 'reads': 62, 'readers_count': 61, 'score': 117.4, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 4}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/14', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 4}], 'current_user_reaction': None, 'reaction_users_count': 4, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 94201, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-10-13T00:39:20.381Z', 'cooked': 'Apologies for the interruption, it should be back to normal now.
', 'post_number': 15, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-13T00:39:20.381Z', 'reply_count': 0, 'reply_to_post_number': 14, 'quote_count': 0, 'incoming_link_count': 10, 'reads': 49, 'readers_count': 48, 'score': 104.8, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/15', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 94234, 'name': 'Sanjana K', 'username': 'SanjanaKannan', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/ce7236/{size}.png', 'created_at': '2023-10-13T06:59:25.130Z', 'cooked': '@radames any idea by when it will be back to normal? I’m still facing the error
', 'post_number': 16, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-13T06:59:25.130Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 47, 'readers_count': 46, 'score': 24.4, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Sanjana K', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28627, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/16', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 94436, 'name': 'Dan Moen', 'username': '155elkhorn', 'avatar_template': '/user_avatar/discuss.huggingface.co/155elkhorn/{size}/19313_2.png', 'created_at': '2023-10-14T15:11:02.165Z', 'cooked': 'Spaces would not start for me this morning, but after factory resets they are running.
', 'post_number': 17, 'post_type': 1, 'posts_count': 24, 'updated_at': '2023-10-14T15:11:02.165Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 43, 'readers_count': 42, 'score': 88.6, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Dan Moen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 28476, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/17', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 152003, 'name': 'Jose Benitez', 'username': 'joselobenitezg', 'avatar_template': '/user_avatar/discuss.huggingface.co/joselobenitezg/{size}/22024_2.png', 'created_at': '2024-08-27T06:12:23.257Z', 'cooked': 'I have the same situation right now! ZeroGPU just freeze in ‘Running’
', 'post_number': 18, 'post_type': 1, 'posts_count': 24, 'updated_at': '2024-08-27T06:12:23.257Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 15, 'readers_count': 14, 'score': 13.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Jose Benitez', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 35634, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/18', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 152004, 'name': 'Jose Benitez', 'username': 'joselobenitezg', 'avatar_template': '/user_avatar/discuss.huggingface.co/joselobenitezg/{size}/22024_2.png', 'created_at': '2024-08-27T06:17:21.051Z', 'cooked': 'stuck in last commit Sapiens Demo - a Hugging Face Space by joselobenitezg
', 'post_number': 19, 'post_type': 1, 'posts_count': 24, 'updated_at': '2024-08-27T06:17:21.051Z', 'reply_count': 0, 'reply_to_post_number': 18, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 17, 'readers_count': 16, 'score': 3.4, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Jose Benitez', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/joselobenitezg/sapiens-demo', 'internal': False, 'reflection': False, 'title': 'Sapiens Demo - a Hugging Face Space by joselobenitezg', 'clicks': 9}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 35634, 'username': 'joselobenitezg', 'name': 'Jose Benitez', 'avatar_template': '/user_avatar/discuss.huggingface.co/joselobenitezg/{size}/22024_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 35634, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/19', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 152127, 'name': 'Jose Benitez', 'username': 'joselobenitezg', 'avatar_template': '/user_avatar/discuss.huggingface.co/joselobenitezg/{size}/22024_2.png', 'created_at': '2024-08-27T18:09:49.244Z', 'cooked': '@julien-c any idea?
', 'post_number': 20, 'post_type': 1, 'posts_count': 24, 'updated_at': '2024-08-27T18:09:49.244Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 15, 'readers_count': 14, 'score': 23.0, 'yours': False, 'topic_id': 54149, 'topic_slug': 'space-wont-start-logs-not-found', 'display_username': 'Jose Benitez', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 35634, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/space-wont-start-logs-not-found/54149/20', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Here’s the error I’m seeing for Container logs:
+Error: Failed to load logs: Not Found. Logs are persisted for 30 days after the Space stops running.
","Apologies for the interruption, it should be back to normal now.
" +Why is Static Cache latency high?,https://discuss.huggingface.co/t/why-is-static-cache-latency-high/157280,157280,9,2025-05-29 16:11:44.321000+00:00,"[{'id': 224686, 'name': 'Yuyao Huang', 'username': 'exhyy', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/977dab/{size}.png', 'created_at': '2025-05-29T16:11:44.386Z', 'cooked': '\n\nIn the above document, “Static Cache” is marked as having high latency. I’m finding this a bit counterintuitive. My understanding is that a Static Cache, by pre-allocating memory for the cache, should help avoid dynamic memory allocation during inference. This, in turn, should theoretically lead to a reduction in latency. Am I misunderstanding its implementation or the definition of “latency” in the document?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-29T16:11:44.386Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 30, 'reads': 4, 'readers_count': 3, 'score': 165.8, 'yours': False, 'topic_id': 157280, 'topic_slug': 'why-is-static-cache-latency-high', 'display_username': 'Yuyao Huang', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/en/kv_cache', 'internal': False, 'reflection': False, 'title': 'KV cache strategies', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95473, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/why-is-static-cache-latency-high/157280/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 224697, 'name': 'Riley Fox', 'username': 'Mdrnfox', 'avatar_template': '/user_avatar/discuss.huggingface.co/mdrnfox/{size}/47695_2.png', 'created_at': '2025-05-29T16:45:50.724Z', 'cooked': '\nThis is how I interpreted it. Hugging Face docs says that Static Cache has “High” latency, it isn’t opposing the fact that pre-allocating memory can avoid dynamic allocations—instead, it’s telling you how fast generation runs by default, without any extra steps.
\nHope this helps
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-30T08:01:14.932Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 157280, 'topic_slug': 'why-is-static-cache-latency-high', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/why-is-static-cache-latency-high/157280/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]"," ++In the above document, “Static Cache” is marked as having high latency. I’m finding this a bit counterintuitive. My understanding is that a Static Cache, by pre-allocating memory for the cache, should help avoid dynamic memory allocation during inference. This, in turn, should theoretically lead to a reduction in latency. Am I misunderstanding its implementation or the definition of “latency” in the document?
"," +This is how I interpreted it. Hugging Face docs says that Static Cache has “High” latency, it isn’t opposing the fact that pre-allocating memory can avoid dynamic allocations—instead, it’s telling you how fast generation runs by default, without any extra steps.
+Hope this helps
Hello there,
\nSo i’m working on a ZeroGPU space, and i was able to generate some images out of it.
\nTho after a day, i wanted to share it with some friends and they are not able to generate (they are not logged, no the quota is not full, i also tried without login and had the same issue).
\nHere is the failed logs :
\n2025-05-13 13:50:08 - httpx - INFO - HTTP Request: POST http://device-api.zero/schedule?cgroupPath=%2Fkubepods.slice%2Fkubepods-burstable.slice%2Fkubepods-burstable-pod53d91e08_ca6f_4829_acd7_772d9f243c8d.slice%2Fcri-containerd-04c1f2c1ffa380d58455444191199b49c387cc8223de321c2ba7806ab5afb790.scope&taskId=140013534102432&enableQueue=true&tokenVersion=1&token=<hidden> ""HTTP/1.1 200 OK""\n2025-05-13 13:50:08 - httpx - INFO - HTTP Request: POST http://device-api.zero/allow?allowToken=30dde4f1969ce8a8e2506e28f806789a21b5458a9e8618389a54bb0f851483b7&pid=4746 ""HTTP/1.1 200 OK""\n2025-05-13 13:50:08 - httpx - INFO - HTTP Request: POST http://device-api.zero/release?allowToken=30dde4f1969ce8a8e2506e28f806789a21b5458a9e8618389a54bb0f851483b7&fail=true ""HTTP/1.1 200 OK""\nTraceback (most recent call last):\n File ""/usr/local/lib/python3.10/site-packages/spaces/zero/wrappers.py"", line 140, in worker_init\n torch.init(nvidia_uuid)\n File ""/usr/local/lib/python3.10/site-packages/spaces/zero/torch/patching.py"", line 373, in init\n torch.Tensor([0]).cuda()\n File ""/usr/local/lib/python3.10/site-packages/torch/cuda/__init__.py"", line 319, in _lazy_init\n torch._C._cuda_init()\nRuntimeError: No CUDA GPUs are available\n\nTraceback (most recent call last):\n File ""/usr/local/lib/python3.10/site-packages/gradio/queueing.py"", line 536, in process_events\n response = await route_utils.call_process_api(\n File ""/usr/local/lib/python3.10/site-packages/gradio/route_utils.py"", line 322, in call_process_api\n output = await app.get_blocks().process_api(\n File ""/usr/local/lib/python3.10/site-packages/gradio/blocks.py"", line 1935, in process_api\n result = await self.call_function(\n File ""/usr/local/lib/python3.10/site-packages/gradio/blocks.py"", line 1520, in call_function\n prediction = await anyio.to_thread.run_sync( # type: ignore\n File ""/usr/local/lib/python3.10/site-packages/anyio/to_thread.py"", line 56, in run_sync\n return await get_async_backend().run_sync_in_worker_thread(\n File ""/usr/local/lib/python3.10/site-packages/anyio/_backends/_asyncio.py"", line 2470, in run_sync_in_worker_thread\n return await future\n File ""/usr/local/lib/python3.10/site-packages/anyio/_backends/_asyncio.py"", line 967, in run\n result = context.run(func, *args)\n File ""/usr/local/lib/python3.10/site-packages/gradio/utils.py"", line 826, in wrapper\n response = f(*args, **kwargs)\n File ""/usr/local/lib/python3.10/site-packages/gradio/utils.py"", line 826, in wrapper\n response = f(*args, **kwargs)\n File ""/usr/local/lib/python3.10/site-packages/spaces/zero/wrappers.py"", line 214, in gradio_handler\n raise error(""ZeroGPU worker error"", res.error_cls)\ngradio.exceptions.Error: \'RuntimeError\'\nand a working one :
\n2025-05-13 13:40:38 - httpx - INFO - HTTP Request: POST http://device-api.zero/schedule?cgroupPath=%2Fkubepods.slice%2Fkubepods-burstable.slice%2Fkubepods-burstable-pod53d91e08_ca6f_4829_acd7_772d9f243c8d.slice%2Fcri-containerd-04c1f2c1ffa380d58455444191199b49c387cc8223de321c2ba7806ab5afb790.scope&taskId=140013534102432&enableQueue=true&tokenVersion=1&token=<hidden> ""HTTP/1.1 200 OK""\n2025-05-13 13:40:38 - httpx - INFO - HTTP Request: POST http://device-api.zero/allow?allowToken=da5eb1a48aafb766ccf710678d8812ca135ce74d51e310832bb0a7da156dd51f&pid=4523 ""HTTP/1.1 200 OK""\n2025-05-13 13:40:41 - __main__ - INFO - Starting generation with parameters: {\n ""prompt"": ""masterpiece, best quality, amazing quality, 1girl"",\n ""negative_prompt"": ""sensitive, nsfw, explicit, bad quality, worst quality, worst detail, sketch, censor"",\n ""resolution"": ""1248 x 1824"",\n ""guidance_scale"": 7,\n ""num_inference_steps"": 28,\n ""seed"": 1857728698,\n ""sampler"": ""Euler a"",\n ""use_upscaler"": null\n}\n2025-05-13 13:40:49 - __main__ - INFO - Image 1/1 saved as ./outputs/20584bdd-e9bc-4691-8399-7bb96e8dcf7b.png\n2025-05-13 13:40:49 - __main__ - INFO - Generation completed successfully in 8.03 seconds\n2025-05-13 13:40:49 - httpx - INFO - HTTP Request: POST http://device-api.zero/release?allowToken=da5eb1a48aafb766ccf710678d8812ca135ce74d51e310832bb0a7da156dd51f&fail=false ""HTTP/1.1 200 OK""\nYes, the import spaces is at the top.
\nNo i’m not using weird pipelines, just “lpw_stable_diffusion_xl” copied from the repo to work with “from_single file”
Just after sending the message, i got the no GPU also on my account.
\nAnd right now, it seems to be woking again both with and without account.
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-13T13:12:43.972Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 17, 'readers_count': 16, 'score': 63.4, 'yours': False, 'topic_id': 154885, 'topic_slug': 'zerogpu-space-no-cuda-gpus-are-available', 'display_username': 'Ibaraki Douji', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93790, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/zerogpu-space-no-cuda-gpus-are-available/154885/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221725, 'name': 'Ibaraki Douji', 'username': 'IbarakiDouji', 'avatar_template': '/user_avatar/discuss.huggingface.co/ibarakidouji/{size}/47435_2.png', 'created_at': '2025-05-13T19:31:45.960Z', 'cooked': 'After more time it happen again.
\nMaybe it’s just there is too much ZeroGPU spaces used at the time.
\nJust hope that someone can clarify the real cause of it.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-13T19:31:45.960Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 16, 'readers_count': 15, 'score': 38.2, 'yours': False, 'topic_id': 154885, 'topic_slug': 'zerogpu-space-no-cuda-gpus-are-available', 'display_username': 'Ibaraki Douji', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93790, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/zerogpu-space-no-cuda-gpus-are-available/154885/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221752, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-14T02:44:00.213Z', 'cooked': 'After replicating it, it seems to work fine now. It probably just comes and goes.
\nThe Zero GPU has just been replaced, so there might be a bug, so I’ll ping it just to be safe. @hysts @michellehbn
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-14T02:44:00.213Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 15, 'reads': 14, 'readers_count': 13, 'score': 122.8, 'yours': False, 'topic_id': 154885, 'topic_slug': 'zerogpu-space-no-cuda-gpus-are-available', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/zerogpu-space-no-cuda-gpus-are-available/154885/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 224277, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-27T09:30:20.561Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-27T09:30:20.561Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 11.0, 'yours': False, 'topic_id': 154885, 'topic_slug': 'zerogpu-space-no-cuda-gpus-are-available', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/zerogpu-space-no-cuda-gpus-are-available/154885/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello there,
+So i’m working on a ZeroGPU space, and i was able to generate some images out of it.
+Tho after a day, i wanted to share it with some friends and they are not able to generate (they are not logged, no the quota is not full, i also tried without login and had the same issue).
+Here is the failed logs :
+2025-05-13 13:50:08 - httpx - INFO - HTTP Request: POST http://device-api.zero/schedule?cgroupPath=%2Fkubepods.slice%2Fkubepods-burstable.slice%2Fkubepods-burstable-pod53d91e08_ca6f_4829_acd7_772d9f243c8d.slice%2Fcri-containerd-04c1f2c1ffa380d58455444191199b49c387cc8223de321c2ba7806ab5afb790.scope&taskId=140013534102432&enableQueue=true&tokenVersion=1&token=<hidden> ""HTTP/1.1 200 OK""
+2025-05-13 13:50:08 - httpx - INFO - HTTP Request: POST http://device-api.zero/allow?allowToken=30dde4f1969ce8a8e2506e28f806789a21b5458a9e8618389a54bb0f851483b7&pid=4746 ""HTTP/1.1 200 OK""
+2025-05-13 13:50:08 - httpx - INFO - HTTP Request: POST http://device-api.zero/release?allowToken=30dde4f1969ce8a8e2506e28f806789a21b5458a9e8618389a54bb0f851483b7&fail=true ""HTTP/1.1 200 OK""
+Traceback (most recent call last):
+ File ""/usr/local/lib/python3.10/site-packages/spaces/zero/wrappers.py"", line 140, in worker_init
+ torch.init(nvidia_uuid)
+ File ""/usr/local/lib/python3.10/site-packages/spaces/zero/torch/patching.py"", line 373, in init
+ torch.Tensor([0]).cuda()
+ File ""/usr/local/lib/python3.10/site-packages/torch/cuda/__init__.py"", line 319, in _lazy_init
+ torch._C._cuda_init()
+RuntimeError: No CUDA GPUs are available
+
+Traceback (most recent call last):
+ File ""/usr/local/lib/python3.10/site-packages/gradio/queueing.py"", line 536, in process_events
+ response = await route_utils.call_process_api(
+ File ""/usr/local/lib/python3.10/site-packages/gradio/route_utils.py"", line 322, in call_process_api
+ output = await app.get_blocks().process_api(
+ File ""/usr/local/lib/python3.10/site-packages/gradio/blocks.py"", line 1935, in process_api
+ result = await self.call_function(
+ File ""/usr/local/lib/python3.10/site-packages/gradio/blocks.py"", line 1520, in call_function
+ prediction = await anyio.to_thread.run_sync( # type: ignore
+ File ""/usr/local/lib/python3.10/site-packages/anyio/to_thread.py"", line 56, in run_sync
+ return await get_async_backend().run_sync_in_worker_thread(
+ File ""/usr/local/lib/python3.10/site-packages/anyio/_backends/_asyncio.py"", line 2470, in run_sync_in_worker_thread
+ return await future
+ File ""/usr/local/lib/python3.10/site-packages/anyio/_backends/_asyncio.py"", line 967, in run
+ result = context.run(func, *args)
+ File ""/usr/local/lib/python3.10/site-packages/gradio/utils.py"", line 826, in wrapper
+ response = f(*args, **kwargs)
+ File ""/usr/local/lib/python3.10/site-packages/gradio/utils.py"", line 826, in wrapper
+ response = f(*args, **kwargs)
+ File ""/usr/local/lib/python3.10/site-packages/spaces/zero/wrappers.py"", line 214, in gradio_handler
+ raise error(""ZeroGPU worker error"", res.error_cls)
+gradio.exceptions.Error: 'RuntimeError'
+and a working one :
+2025-05-13 13:40:38 - httpx - INFO - HTTP Request: POST http://device-api.zero/schedule?cgroupPath=%2Fkubepods.slice%2Fkubepods-burstable.slice%2Fkubepods-burstable-pod53d91e08_ca6f_4829_acd7_772d9f243c8d.slice%2Fcri-containerd-04c1f2c1ffa380d58455444191199b49c387cc8223de321c2ba7806ab5afb790.scope&taskId=140013534102432&enableQueue=true&tokenVersion=1&token=<hidden> ""HTTP/1.1 200 OK""
+2025-05-13 13:40:38 - httpx - INFO - HTTP Request: POST http://device-api.zero/allow?allowToken=da5eb1a48aafb766ccf710678d8812ca135ce74d51e310832bb0a7da156dd51f&pid=4523 ""HTTP/1.1 200 OK""
+2025-05-13 13:40:41 - __main__ - INFO - Starting generation with parameters: {
+ ""prompt"": ""masterpiece, best quality, amazing quality, 1girl"",
+ ""negative_prompt"": ""sensitive, nsfw, explicit, bad quality, worst quality, worst detail, sketch, censor"",
+ ""resolution"": ""1248 x 1824"",
+ ""guidance_scale"": 7,
+ ""num_inference_steps"": 28,
+ ""seed"": 1857728698,
+ ""sampler"": ""Euler a"",
+ ""use_upscaler"": null
+}
+2025-05-13 13:40:49 - __main__ - INFO - Image 1/1 saved as ./outputs/20584bdd-e9bc-4691-8399-7bb96e8dcf7b.png
+2025-05-13 13:40:49 - __main__ - INFO - Generation completed successfully in 8.03 seconds
+2025-05-13 13:40:49 - httpx - INFO - HTTP Request: POST http://device-api.zero/release?allowToken=da5eb1a48aafb766ccf710678d8812ca135ce74d51e310832bb0a7da156dd51f&fail=false ""HTTP/1.1 200 OK""
+Yes, the import spaces is at the top.
+No i’m not using weird pipelines, just “lpw_stable_diffusion_xl” copied from the repo to work with “from_single file”
After replicating it, it seems to work fine now. It probably just comes and goes.
+The Zero GPU has just been replaced, so there might be a bug, so I’ll ping it just to be safe. @hysts @michellehbn
" +Building something that help people who really need help using ai,https://discuss.huggingface.co/t/building-something-that-help-people-who-really-need-help-using-ai/154301,154301,9,2025-05-09 14:15:08.458000+00:00,"[{'id': 220825, 'name': 'Adnan Ahamed Farooqui', 'username': 'adnanahmedfarooqui', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/71c47a/{size}.png', 'created_at': '2025-05-09T14:15:08.520Z', 'cooked': 'I want to make something like that using AI automation and other tools that will help different kinds of people.
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-09T14:15:08.520Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 10, 'readers_count': 9, 'score': 47.0, 'yours': False, 'topic_id': 154301, 'topic_slug': 'building-something-that-help-people-who-really-need-help-using-ai', 'display_username': 'Adnan Ahamed Farooqui', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90632, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/building-something-that-help-people-who-really-need-help-using-ai/154301/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 220827, 'name': 'Tonni Alex', 'username': 'tonnii', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/t/a9adbd/{size}.png', 'created_at': '2025-05-09T14:19:57.020Z', 'cooked': 'That is a great idea. If you want to build something using AI automation and other tools to help different kinds of people, begin by deciding what problem you want to solve and who will use it. Once you know that, choose the right tools such as chatbots, automation platforms, or voice assistants, based on what is needed. Many tools are easy to use and do not require heavy coding. Build one small part at a time, test it with real users, and make sure it is simple and helpful for the people you want to support.
', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-09T14:19:57.164Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 10, 'readers_count': 9, 'score': 32.0, 'yours': False, 'topic_id': 154301, 'topic_slug': 'building-something-that-help-people-who-really-need-help-using-ai', 'display_username': 'Tonni Alex', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93030, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': 'Automatically removed quote of whole previous post.', 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/building-something-that-help-people-who-really-need-help-using-ai/154301/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221050, 'name': 'Adnan Ahamed Farooqui', 'username': 'adnanahmedfarooqui', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/71c47a/{size}.png', 'created_at': '2025-05-10T17:15:39.124Z', 'cooked': 'I am thinking of creating an AI technology that will help in the indoor mapping of different places, fully descriptive, which will help old age people and differently abled people to access those places easily. Can anyone help me with that
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-10T17:15:39.124Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 154301, 'topic_slug': 'building-something-that-help-people-who-really-need-help-using-ai', 'display_username': 'Adnan Ahamed Farooqui', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90632, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/building-something-that-help-people-who-really-need-help-using-ai/154301/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221201, 'name': 'Mahmut C', 'username': 'mahmutc', 'avatar_template': '/user_avatar/discuss.huggingface.co/mahmutc/{size}/52583_2.png', 'created_at': '2025-05-11T13:30:21.276Z', 'cooked': '\nDo you think something like this?
\nUser: “Take me to the cardiology wing.”
\nAI Response: “You are 20 meters from the elevator. Take the elevator to the second floor. Upon exit, turn left and follow the tactile floor markings. A staff help desk will be on your right in 30 meters.”
Yess exactly like this …can make further changes by getting user input that will help people to navigate the places easily…also in our map we can mark places that is fully accessible partially accessable and not accessible in outdoor map…
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-12T07:27:14.582Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 15.8, 'yours': False, 'topic_id': 154301, 'topic_slug': 'building-something-that-help-people-who-really-need-help-using-ai', 'display_username': 'Adnan Ahamed Farooqui', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 61570, 'username': 'mahmutc', 'name': 'Mahmut C', 'avatar_template': '/user_avatar/discuss.huggingface.co/mahmutc/{size}/52583_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90632, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/building-something-that-help-people-who-really-need-help-using-ai/154301/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 224274, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-27T09:00:06.119Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-05-27T09:00:06.119Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 154301, 'topic_slug': 'building-something-that-help-people-who-really-need-help-using-ai', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/building-something-that-help-people-who-really-need-help-using-ai/154301/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]",I want to make something like that using AI automation and other tools that will help different kinds of people.
,Yess exactly like this …can make further changes by getting user input that will help people to navigate the places easily…also in our map we can mark places that is fully accessible partially accessable and not accessible in outdoor map…
+Optimal Approach for Fine-Tuning LayoutLMv3 for Token Classification with 80 Labels,https://discuss.huggingface.co/t/optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels/156857,156857,13,2025-05-26 11:29:11.157000+00:00,"[{'id': 224129, 'name': 'hugo pavy', 'username': 'hugobee', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugobee/{size}/48285_2.png', 'created_at': '2025-05-26T11:29:11.235Z', 'cooked': 'Hello everyone,
\nI’m trying to extract medical information from PDF files using LayoutLMv3 for token classification.
\nI’ve successfully fine-tuned the model for a few different kinds of tokens (name, date of birth, patient ID, etc.), but now I want to scale up to around 80 different labels.
\nI’m wondering if it’s better to train one model for all labels or to decompose the task into multiple specialized models (like just models of around 10 labels). Any advice or experiences would be greatly appreciated!
\nHas anyone encountered a similar issue or have any advice on the best approach? Thanks in advance for your help!
\nHave a good day,
\nHugo
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-26T11:29:11.235Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 63, 'reads': 8, 'readers_count': 7, 'score': 286.6, 'yours': False, 'topic_id': 156857, 'topic_slug': 'optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels', 'display_username': 'hugo pavy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95134, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels/156857/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 224136, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-26T13:13:15.723Z', 'cooked': '\n\nif it’s better to train one model for all labels or to decompose the task into multiple specialized models (like just models of around 10 labels)
\n
Looking at the dataset used to train LayoutLMv2, it seems that a number of items within 20 is more appropriate. I think v3 probably has similar characteristics.
\n\nWell, small models are often not suitable for processing many items at once, so it is safer to divide them into multiple models. Even if you continue to train a single model, it is a good idea to save the current successful weights somewhere.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-26T13:13:15.723Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 21.2, 'yours': False, 'topic_id': 156857, 'topic_slug': 'optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/FrancophonIA/XFUND', 'internal': False, 'reflection': False, 'title': 'FrancophonIA/XFUND · Datasets at Hugging Face', 'clicks': 5}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels/156857/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 224149, 'name': 'hugo pavy', 'username': 'hugobee', 'avatar_template': '/user_avatar/discuss.huggingface.co/hugobee/{size}/48285_2.png', 'created_at': '2025-05-26T14:57:05.139Z', 'cooked': 'Thanks you for your response! I’m gonna try that
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-26T14:57:05.139Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 156857, 'topic_slug': 'optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels', 'display_username': 'hugo pavy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 95134, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels/156857/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 224270, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-27T08:08:12.063Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-05-27T08:08:12.063Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 156857, 'topic_slug': 'optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/optimal-approach-for-fine-tuning-layoutlmv3-for-token-classification-with-80-labels/156857/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone,
+I’m trying to extract medical information from PDF files using LayoutLMv3 for token classification.
+I’ve successfully fine-tuned the model for a few different kinds of tokens (name, date of birth, patient ID, etc.), but now I want to scale up to around 80 different labels.
+I’m wondering if it’s better to train one model for all labels or to decompose the task into multiple specialized models (like just models of around 10 labels). Any advice or experiences would be greatly appreciated!
+Has anyone encountered a similar issue or have any advice on the best approach? Thanks in advance for your help!
+Have a good day,
+Hugo
","++if it’s better to train one model for all labels or to decompose the task into multiple specialized models (like just models of around 10 labels)
+
Looking at the dataset used to train LayoutLMv2, it seems that a number of items within 20 is more appropriate. I think v3 probably has similar characteristics.
+ +Well, small models are often not suitable for processing many items at once, so it is safer to divide them into multiple models. Even if you continue to train a single model, it is a good idea to save the current successful weights somewhere.
" +Need help to find old Embeddings I lost during PC installation,https://discuss.huggingface.co/t/need-help-to-find-old-embeddings-i-lost-during-pc-installation/156873,156873,13,2025-05-26 14:26:01.784000+00:00,"[{'id': 224147, 'name': 'Mary', 'username': 'fantasy-mary', 'avatar_template': '/user_avatar/discuss.huggingface.co/fantasy-mary/{size}/48307_2.png', 'created_at': '2025-05-26T14:26:01.849Z', 'cooked': 'Hi everyone,
\nI am looking for help, I used some embeddings but after I reinstalled Windows to my PC I lost my StableDiffusion folder. Now I reinstalled StableDiffusion but I can’t find all embeddings.
\nThe specific embeddings I am looking for are called “fFaceDetail, SkinHairDetail, EyeDetail, OverallDetail and SkinDetailNeg-neg”. I did not rename them, I am 100% sure they are from civitai and all from one creator but I can’t find them there anymore.
\nMaybe someone knows them, knows where I can find them or even got them by themself and are willing to share them.
\nThanks in advance
Hi @fantasy-mary, it’s a shame you lost your data
\nI found this while searching the web. I hope it helps!
Adrian Araya
\nMachine Learning Engineer at RidgeRun.ai
\nContact us: support@ridgerun.ai
Oh my god you are great, thank you !!
\nI searched for it the whole day and could not find them.
I’m glad it worked for you, have a nice day!
\nAdrian Araya
\nMachine Learning Engineer at RidgeRun.ai
\nContact us: support@ridgerun.ai
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-27T04:43:22.509Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 156873, 'topic_slug': 'need-help-to-find-old-embeddings-i-lost-during-pc-installation', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/need-help-to-find-old-embeddings-i-lost-during-pc-installation/156873/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi everyone,
+I am looking for help, I used some embeddings but after I reinstalled Windows to my PC I lost my StableDiffusion folder. Now I reinstalled StableDiffusion but I can’t find all embeddings.
+The specific embeddings I am looking for are called “fFaceDetail, SkinHairDetail, EyeDetail, OverallDetail and SkinDetailNeg-neg”. I did not rename them, I am 100% sure they are from civitai and all from one creator but I can’t find them there anymore.
+Maybe someone knows them, knows where I can find them or even got them by themself and are willing to share them.
+Thanks in advance
Hi @fantasy-mary, it’s a shame you lost your data
+I found this while searching the web. I hope it helps!
Adrian Araya
+Machine Learning Engineer at RidgeRun.ai
+Contact us: support@ridgerun.ai
Hello everyone😊,
\nI’d like to test the model on the free CPU environment—do you have any suggestions?
I’m encountering an error when trying to deploy the Qwen1.5-0.5B-Chat model in my Hugging Face Space running on CPU-only (free) .
\nMyQwen1.5 0.5B Chat - a Hugging Face Space by funme
\nThank you
\nHere the full log: tokenizer_config.json: 0%| | 0.00/1.29k [00:00<?, ?B/s]
\ntokenizer_config.json: 100%|██████████| 1.29k/1.29k [00:00<00:00, 7.24MB/s]
\nvocab.json: 0%| | 0.00/2.78M [00:00<?, ?B/s]
\nvocab.json: 100%|██████████| 2.78M/2.78M [00:00<00:00, 27.1MB/s]
\nmerges.txt: 0%| | 0.00/1.67M [00:00<?, ?B/s]
\nmerges.txt: 100%|██████████| 1.67M/1.67M [00:00<00:00, 31.1MB/s]
\ntokenizer.json: 0%| | 0.00/7.03M [00:00<?, ?B/s]
\ntokenizer.json: 100%|██████████| 7.03M/7.03M [00:00<00:00, 58.3MB/s]
\nconfig.json: 0%| | 0.00/1.26k [00:00<?, ?B/s]
\nconfig.json: 100%|██████████| 1.26k/1.26k [00:00<00:00, 7.28MB/s]
\nTraceback (most recent call last):
\nFile “/home/user/app/app.py”, line 9, in
\nmodel = AutoModelForCausalLM.from_pretrained(
\nFile “/usr/local/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py”, line 571, in from_pretrained
\nreturn model_class.from_pretrained(
\nFile “/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py”, line 309, in _wrapper
\nreturn func(*args, **kwargs)
\nFile “/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py”, line 4389, in from_pretrained
\nhf_quantizer.validate_environment(
\nFile “/usr/local/lib/python3.10/site-packages/transformers/quantizers/quantizer_gptq.py”, line 65, in validate_environment
\nraise RuntimeError(“GPU is required to quantize or run quantize model.”)
\nRuntimeError: GPU is required to quantize or run quantize model.
It may be possible to use a quantized model in a CPU environment, but it would probably be faster to simply use a non-quantized model in this case.
\n#MODEL_ID = ""Qwen/Qwen1.5-0.5B-Chat-GPTQ-Int4""\nMODEL_ID = ""Qwen/Qwen1.5-0.5B-Chat""\nThank you😊 , I need a model size smaller than 700 MB , I’m going to change model, if I can’t use this model
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-23T16:04:58.404Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 1, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 156535, 'topic_slug': 'runtimeerror-gpu-is-required-to-quantize-or-run-quantize-model-qwen1-5-0-5b-chat-in-my-space', 'display_username': ""I'm cute"", 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94919, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/runtimeerror-gpu-is-required-to-quantize-or-run-quantize-model-qwen1-5-0-5b-chat-in-my-space/156535/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 223783, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-24T04:05:31.298Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-05-24T04:05:31.298Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 156535, 'topic_slug': 'runtimeerror-gpu-is-required-to-quantize-or-run-quantize-model-qwen1-5-0-5b-chat-in-my-space', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/runtimeerror-gpu-is-required-to-quantize-or-run-quantize-model-qwen1-5-0-5b-chat-in-my-space/156535/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone😊,
+I’d like to test the model on the free CPU environment—do you have any suggestions?
I’m encountering an error when trying to deploy the Qwen1.5-0.5B-Chat model in my Hugging Face Space running on CPU-only (free) .
+MyQwen1.5 0.5B Chat - a Hugging Face Space by funme
+Thank you
+Here the full log: tokenizer_config.json: 0%| | 0.00/1.29k [00:00<?, ?B/s]
+tokenizer_config.json: 100%|██████████| 1.29k/1.29k [00:00<00:00, 7.24MB/s]
+vocab.json: 0%| | 0.00/2.78M [00:00<?, ?B/s]
+vocab.json: 100%|██████████| 2.78M/2.78M [00:00<00:00, 27.1MB/s]
+merges.txt: 0%| | 0.00/1.67M [00:00<?, ?B/s]
+merges.txt: 100%|██████████| 1.67M/1.67M [00:00<00:00, 31.1MB/s]
+tokenizer.json: 0%| | 0.00/7.03M [00:00<?, ?B/s]
+tokenizer.json: 100%|██████████| 7.03M/7.03M [00:00<00:00, 58.3MB/s]
+config.json: 0%| | 0.00/1.26k [00:00<?, ?B/s]
+config.json: 100%|██████████| 1.26k/1.26k [00:00<00:00, 7.28MB/s]
+Traceback (most recent call last):
+File “/home/user/app/app.py”, line 9, in
+model = AutoModelForCausalLM.from_pretrained(
+File “/usr/local/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py”, line 571, in from_pretrained
+return model_class.from_pretrained(
+File “/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py”, line 309, in _wrapper
+return func(*args, **kwargs)
+File “/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py”, line 4389, in from_pretrained
+hf_quantizer.validate_environment(
+File “/usr/local/lib/python3.10/site-packages/transformers/quantizers/quantizer_gptq.py”, line 65, in validate_environment
+raise RuntimeError(“GPU is required to quantize or run quantize model.”)
+RuntimeError: GPU is required to quantize or run quantize model.
Thank you😊 , I need a model size smaller than 700 MB , I’m going to change model, if I can’t use this model
" +"Configuration error, deleted readme.md",https://discuss.huggingface.co/t/configuration-error-deleted-readme-md/39258,39258,24,2023-05-09 12:39:22.525000+00:00,"[{'id': 68623, 'name': 'Javed', 'username': 'JavedA', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/3bc359/{size}.png', 'created_at': '2023-05-09T12:39:22.584Z', 'cooked': 'Hi, I deleted my README.md pushed it and when I created a new one, pushing it won’t work.
\nThe repo is: Master Thesis - a Hugging Face Space by JavedA
It tells me that there is a configuration error. However, I cannot create a README, neither locally to push it nor using the web view.
\nThank you for your time and effort
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2023-05-09T12:39:53.309Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 725, 'reads': 27, 'readers_count': 26, 'score': 3565.4, 'yours': False, 'topic_id': 39258, 'topic_slug': 'configuration-error-deleted-readme-md', 'display_username': 'Javed', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/JavedA/master_Thesis', 'internal': False, 'reflection': False, 'title': 'Master Thesis - a Hugging Face Space by JavedA', 'clicks': 5}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 18152, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/configuration-error-deleted-readme-md/39258/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 68625, 'name': 'Javed', 'username': 'JavedA', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/3bc359/{size}.png', 'created_at': '2023-05-09T12:54:14.652Z', 'cooked': 'The issue could be solved - I do not know why it worked this time. I just copied the README from a test space and inserted it. Maybe the additional: Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference solved the issue.
Anyhow, the issue could be resolved by simply using the following content for the readme.md
\n\n---\ntitle: Test\nemoji: ⚡\ncolorFrom: pink\ncolorTo: blue\nsdk: static\npinned: false\n---\n\nCheck out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference\nFor me it was also making sure of the filename case. README.md.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2024-11-24T05:30:03.457Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 13, 'readers_count': 12, 'score': 42.6, 'yours': False, 'topic_id': 39258, 'topic_slug': 'configuration-error-deleted-readme-md', 'display_username': 'J Blu', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 48868, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/configuration-error-deleted-readme-md/39258/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 223647, 'name': 'Diseph D', 'username': 'sephdev', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/c4cdca/{size}.png', 'created_at': '2025-05-23T06:48:01.080Z', 'cooked': 'Naming the file in all caps solved mine too
', 'post_number': 4, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-23T06:48:39.734Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 39258, 'topic_slug': 'configuration-error-deleted-readme-md', 'display_username': 'Diseph D', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 48868, 'username': 'johnblues', 'name': 'J Blu', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/f475e1/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94869, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/configuration-error-deleted-readme-md/39258/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi, I deleted my README.md pushed it and when I created a new one, pushing it won’t work.
+The repo is: Master Thesis - a Hugging Face Space by JavedA
It tells me that there is a configuration error. However, I cannot create a README, neither locally to push it nor using the web view.
+Thank you for your time and effort
","The issue could be solved - I do not know why it worked this time. I just copied the README from a test space and inserted it. Maybe the additional: Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference solved the issue.
Anyhow, the issue could be resolved by simply using the following content for the readme.md
+
+---
+title: Test
+emoji: ⚡
+colorFrom: pink
+colorTo: blue
+sdk: static
+pinned: false
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Using Trainer, and it appears that if I load any class from accelerate, the Trainer doesn’t perform its accelerate magic behind the scenes, meaning I get an error like this:
\n[rank1]: File ""/opt/code/repos/MyProject/.venv/lib/python3.12/site-packages/transformers/modeling_utils.py"", line 5779, in caching_allocator_warmup\n[rank1]: re.compile(""|"".join([re.escape(plan) for plan in model._tp_plan]))\n[rank1]: ^^^^^^^^^^^^^^\n[rank1]: TypeError: \'NoneType\' object is not iterable\nI have two use cases where I’d like slightly more control:
\nMy script creates a directory with a timestamp, and there is a synchronization issue that creates two checkpoint directories, one for each GPU.
\nI load two models, the second attempt to load it always fails with this error. It appears that once the Trainer/TrainingArguments go out of scope, the accelerate process is torn down and doesn’t get reinitialized.
\nHow can I take more control of the process? Is there a way to manually manage accelerate with the Trainer and TrainingArguments objects? How about synchronization primitives: something that allows a function to run on the main process before forking to the subprocesses? I tried the decorators, but they cause the Trainer code to crash with the same error.
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-22T01:25:41.191Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 46, 'reads': 6, 'readers_count': 5, 'score': 226.0, 'yours': False, 'topic_id': 156255, 'topic_slug': 'synchronizing-state-trainer-and-accelerate', 'display_username': 'Don B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 5859, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/synchronizing-state-trainer-and-accelerate/156255/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 223572, 'name': 'Don B', 'username': 'donb', 'avatar_template': '/user_avatar/discuss.huggingface.co/donb/{size}/3744_2.png', 'created_at': '2025-05-22T16:45:23.597Z', 'cooked': ""I have worked around this issue by modifying caching_allocator_warmup to set the tp_plan_regex to None if in addition to if _torch_distributed_available and torch.distributed.is_initialized() it checks if model._tp_plan is valid:
\nif _torch_distributed_available and torch.distributed.is_initialized() and hasattr(model, '_tp_plan') and model._tp_plan is not None.
This prevents the failure and ddp is working correctly across multiple invocations inside the Trainers.
\nI don’t know the implications of this _tp_plan modification, but my AI pair programmer suggests that when using accelerate launch and ddp, model._tp_plan should be None. (my pair programmer was not helpful in fixing this naturally - no impactful suggestions). If I understood it better I would create an issue and submit a pull request. For now, I will just monkeypatch it.
"", 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-22T16:45:23.597Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 20.8, 'yours': False, 'topic_id': 156255, 'topic_slug': 'synchronizing-state-trainer-and-accelerate', 'display_username': 'Don B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 5859, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/synchronizing-state-trainer-and-accelerate/156255/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 223573, 'name': 'Don B', 'username': 'donb', 'avatar_template': '/user_avatar/discuss.huggingface.co/donb/{size}/3744_2.png', 'created_at': '2025-05-22T16:47:29.131Z', 'cooked': 'Also noting that the few issues I’ve found related to the iteration over a None _tp_plan is the model’s fault and addressable through proper _post_init usage. This seems like a brittle solution and one that won’t scale across all the sources for custom models.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-22T16:47:29.131Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 15.8, 'yours': False, 'topic_id': 156255, 'topic_slug': 'synchronizing-state-trainer-and-accelerate', 'display_username': 'Don B', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 5859, 'username': 'donb', 'name': 'Don B', 'avatar_template': '/user_avatar/discuss.huggingface.co/donb/{size}/3744_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 5859, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/synchronizing-state-trainer-and-accelerate/156255/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 223634, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-23T04:48:23.208Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-05-23T04:48:23.208Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.6, 'yours': False, 'topic_id': 156255, 'topic_slug': 'synchronizing-state-trainer-and-accelerate', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/synchronizing-state-trainer-and-accelerate/156255/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Using Trainer, and it appears that if I load any class from accelerate, the Trainer doesn’t perform its accelerate magic behind the scenes, meaning I get an error like this:
+[rank1]: File ""/opt/code/repos/MyProject/.venv/lib/python3.12/site-packages/transformers/modeling_utils.py"", line 5779, in caching_allocator_warmup
+[rank1]: re.compile(""|"".join([re.escape(plan) for plan in model._tp_plan]))
+[rank1]: ^^^^^^^^^^^^^^
+[rank1]: TypeError: 'NoneType' object is not iterable
+I have two use cases where I’d like slightly more control:
+My script creates a directory with a timestamp, and there is a synchronization issue that creates two checkpoint directories, one for each GPU.
+I load two models, the second attempt to load it always fails with this error. It appears that once the Trainer/TrainingArguments go out of scope, the accelerate process is torn down and doesn’t get reinitialized.
+How can I take more control of the process? Is there a way to manually manage accelerate with the Trainer and TrainingArguments objects? How about synchronization primitives: something that allows a function to run on the main process before forking to the subprocesses? I tried the decorators, but they cause the Trainer code to crash with the same error.
","I have worked around this issue by modifying caching_allocator_warmup to set the tp_plan_regex to None if in addition to if _torch_distributed_available and torch.distributed.is_initialized() it checks if model._tp_plan is valid:
+if _torch_distributed_available and torch.distributed.is_initialized() and hasattr(model, '_tp_plan') and model._tp_plan is not None.
This prevents the failure and ddp is working correctly across multiple invocations inside the Trainers.
+I don’t know the implications of this _tp_plan modification, but my AI pair programmer suggests that when using accelerate launch and ddp, model._tp_plan should be None. (my pair programmer was not helpful in fixing this naturally - no impactful suggestions). If I understood it better I would create an issue and submit a pull request. For now, I will just monkeypatch it.
" +"Can’t upload my model, stuck on “hashing”",https://discuss.huggingface.co/t/cant-upload-my-model-stuck-on-hashing/106539,106539,5,2024-09-13 03:28:43.245000+00:00,"[{'id': 155103, 'name': 'Phoenix Storm Jr.', 'username': 'PhoenixStormJr', 'avatar_template': '/user_avatar/discuss.huggingface.co/phoenixstormjr/{size}/31552_2.png', 'created_at': '2024-09-13T03:28:43.296Z', 'cooked': 'The title says pretty much everything. I was able to upload with a Google Colab hack, but normally, I can’t. I attached the files down below. Can anyone figure out what the deal is?
\nI “fixed” the problem by uploading them with google colab, but I don’t like this solution. Why won’t it upload normally? Here is the colab link:
\n\nHere is the screenshot showing the huggingface refusing to hash:
\nAnd here are the files that wouldn’t hash:
\n\n\nWhat’s going on?
', 'post_number': 1, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T03:28:43.296Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 562, 'reads': 18, 'readers_count': 17, 'score': 2768.6, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://colab.research.google.com/github/PhoenixStormJr/Upload-File-To-Huggingface-With-Google-Colab/blob/main/Upload_File_To_Huggingface.ipynb', 'internal': False, 'reflection': False, 'title': 'Google Colab', 'clicks': 7}, {'url': 'https://huggingface.co/PhoenixStormJr/Megaman-NT-Warrior-Aki-RVC/tree/main', 'internal': False, 'reflection': False, 'title': 'PhoenixStormJr/Megaman-NT-Warrior-Aki-RVC at main', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 155107, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T03:52:10.596Z', 'cooked': 'I was able to upload the file normally with Firfox, am I uploading the wrong file? Is there some kind of weird environment-dependent error?
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T03:52:49.667Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 16, 'readers_count': 15, 'score': 23.2, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/John6666/uploadtest', 'internal': False, 'reflection': False, 'title': 'John6666/uploadtest · Hugging Face', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155108, 'name': 'Phoenix Storm Jr.', 'username': 'PhoenixStormJr', 'avatar_template': '/user_avatar/discuss.huggingface.co/phoenixstormjr/{size}/31552_2.png', 'created_at': '2024-09-13T03:53:58.653Z', 'cooked': 'I tried uploading with a windows virtual machine as well, and with Linux. It used to work but no longer works. This leads me to think there’s a problem on my local computer. However, uploading to google drive works just fine. Any ideas what could be wrong with my computer? I’ve tried google chrome, firefox, chromium, and microsoft edge browsers.
\nYou uploaded the right files. I just don’t get it. It must be a local problem.
', 'post_number': 3, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T03:55:08.732Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 13, 'readers_count': 12, 'score': 17.6, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155109, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T03:58:46.950Z', 'cooked': 'In that case, it’s not your computer, it’s your ISP, or something between the CDN (I don’t know which one) that HF uses and the ISP, or something in that area.
\nBut since we can have a conversation on the HF forum like this, I don’t see how a normal tracert would be able to determine the cause…
\nAnother possibility is that HF’s file system is malfunctioning in some way.
The fact that it’s reproducible is tricky. It’s not a temporary server error.
… uuuh… I don’t think I understood… I mean, I am a beginner and stuff. Basically, I’m getting that I can’t fix it UNLESS I use Google Colab, right?
(I know what an ISP is, like AT&T, but not a CDN)
\n(So… you’re saying my PC is good then, right? It’s a network problem?)
', 'post_number': 5, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T04:07:28.900Z', 'reply_count': 1, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 13, 'readers_count': 12, 'score': 22.6, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155111, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T04:11:03.611Z', 'cooked': 'No, I’m an amateur at networking too!
\nUsing Colab to get around it is half right as long as it works, but something is definitely wrong on the HF side or your side or both.
\nIf I could isolate the problem a bit more, I could send a mentions to the HF staff to let them know, but since I can’t reproduce the problem (if the above can be uploaded, that’s OK, right?) You’re the only one who can verify…
If it’s the same with Linux, it’s hard to imagine, for example, that your PC has been hit by a virus. If your router was attacked by a virus, it might be possible, but I have no experience.
\nIf your hard disk is corrupted, Colab must not be able to help you.
\nIf the problem is upstream of that, you can use a VPN to bypass it, or something like that. (If you can use Colab to get around this, maybe VPN method will work?)
Thanks for your help anyway. I’ll just keep this open and wait to see if anyone else gets this issue. I appreciate your help.
(As for anyone else, who may be experiencing this issue, please comment! I know if it happened to me, it had to of happened to someone else.)
', 'post_number': 7, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T04:17:09.004Z', 'reply_count': 1, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 11, 'readers_count': 10, 'score': 12.2, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155113, 'name': 'Phoenix Storm Jr.', 'username': 'PhoenixStormJr', 'avatar_template': '/user_avatar/discuss.huggingface.co/phoenixstormjr/{size}/31552_2.png', 'created_at': '2024-09-13T04:19:24.177Z', 'cooked': 'So, I tested on my ANDROID Phone, and THAT worked! So I know it’s a problem with my computer specifically. It has to be.
', 'post_number': 8, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T04:19:24.177Z', 'reply_count': 0, 'reply_to_post_number': 6, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 9, 'readers_count': 8, 'score': 1.8, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155114, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T04:19:57.812Z', 'cooked': '\n\nI know if it happened to me, it had to of happened to someone else.
\n
Exactly.
', 'post_number': 9, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T04:19:57.812Z', 'reply_count': 0, 'reply_to_post_number': 7, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 9, 'readers_count': 8, 'score': 6.8, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 64378, 'username': 'PhoenixStormJr', 'name': 'Phoenix Storm Jr.', 'avatar_template': '/user_avatar/discuss.huggingface.co/phoenixstormjr/{size}/31552_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155115, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T04:22:56.799Z', 'cooked': '\n\nSo I know it’s a problem with my computer specifically. It has to be.
\n
Good! (Not good)
\nI wonder what the problem is… is the LAN port broken? Is the cable torn? If you didn’t connect your Android to Wi-Fi and it worked, maybe your ISP is denying access to HF file server?
I have access to every single website on my computer and android. The only difference is huggingface. Both android and my computer are connected to the same wifi network. It’s weird, everything else in my PC is working just great, including online games. Therefore, I know it’s not my ISP.
', 'post_number': 11, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T04:25:24.384Z', 'reply_count': 1, 'reply_to_post_number': 10, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 6.4, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/11', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155118, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T04:30:35.118Z', 'cooked': 'Surely that would mean a PC problem, but what in the world are the possibilities…?
\nIf it’s a hardware problem, online games won’t work, and if it’s a software problem, why not even in a Linux environment?
\nI get it, but there’s more I don’t understand. Well, have you almost succeeded in isolating the problem?
Nope. No idea what now. I just know it’s my own PC that’s the issue. That’s all I know. But it’s not a browser issue since other browsers don’t work either!
', 'post_number': 13, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-13T04:40:54.001Z', 'reply_count': 1, 'reply_to_post_number': 12, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 7, 'readers_count': 6, 'score': 6.4, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/13', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 155134, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-13T06:31:17.979Z', 'cooked': 'I was thinking vaguely about it while working on my own, but I couldn’t come up with anything!
\nIf the PC is also connected via Wi-Fi, the only thing I can think of is that maybe the PC has some special designation in the router settings (you need it sometimes for internet games or something), or maybe the PC’s Wi-Fi adapter is in bad shape or has a bad setting. It’s not impossible, since smartphones are often a newer generation and more powerful when it comes to Wi-Fi.
\nThe easy way to test if this is the cause is to plug the LAN cable from the router directly into the PC, but that’s a pain if you don’t have a cable at home.
Thanks for the advice, but unfortunately it still didn’t work. I plugged in my ethernet cable, and tried uploading, same problem.
\nI think there’s a security issue on Huggingface’s side. Because I can upload to ANY other website just fine. Even my college
\nI made this repository until Huggingface manages to fix the problem:
\n\n', 'post_number': 15, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-29T23:27:06.161Z', 'reply_count': 1, 'reply_to_post_number': 14, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 11.0, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/PhoenixStormJr/Upload-File-To-Huggingface-With-Google-Colab/tree/main', 'internal': False, 'reflection': False, 'title': ""GitHub - PhoenixStormJr/Upload-File-To-Huggingface-With-Google-Colab: Huggingface has a problem with uploading files, so I made this repository to easily upload files. I don't know what the problem with huggingface is. I plan to create a forum to ask for "", 'clicks': 4}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/15', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 158989, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-29T23:38:09.114Z', 'cooked': '\n\nI think there’s a security issue on Huggingface’s side
\n
That’s what I thought, too, but then how does HF pinpoint the restriction to just your PC, even if it’s not intentional?
\nFirst of all, if they’re regulating by account, it shouldn’t even be via Colab.
\nIf they’re regulating by IP, then it wouldn’t work via Android Wi-Fi either.
\nEven the MAC address of the PC changed when you plugged in the ethernet cable, so it’s a bit odd to make this a combined problem with your router. Your router must think your PC is a different person than it was before.
UA may be there because the whole browser industry has changed recently so that it doesn’t change when you change browsers. It does indeed change between Android and PC. But I’ve never heard of pristine IP + UA restrictions in HF.
\nThere was a problem with frequent 500 errors on HF, but it was resolved by the HF staff, so this is probably not the cause of the current problem either.
\n@not-lain @nielsr Do you know anything about it?
', 'post_number': 16, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-09-29T23:38:09.114Z', 'reply_count': 1, 'reply_to_post_number': 15, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 64378, 'username': 'PhoenixStormJr', 'name': 'Phoenix Storm Jr.', 'avatar_template': '/user_avatar/discuss.huggingface.co/phoenixstormjr/{size}/31552_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/16', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 159290, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-10-01T08:57:34.719Z', 'cooked': 'If it’s just one person, you can put it away as a coincidence, but when it’s multiple people, it’s a little suspect. Is it really a problem with the user’s connection?
\n\n\n\n', 'post_number': 17, 'post_type': 1, 'posts_count': 20, 'updated_at': '2024-10-01T08:57:34.719Z', 'reply_count': 1, 'reply_to_post_number': 16, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 5.6, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/SG161222/RealFlux_1.0b_Dev', 'internal': False, 'reflection': False, 'title': 'SG161222/RealFlux_1.0b_Dev · Hugging Face', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/17', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 223193, 'name': 'Phoenix Storm Jr.', 'username': 'PhoenixStormJr', 'avatar_template': '/user_avatar/discuss.huggingface.co/phoenixstormjr/{size}/31552_2.png', 'created_at': '2025-05-20T20:49:22.773Z', 'cooked': 'I encountered a problem with uploading the model to HF (my internet connection has been unstable lately). Once I resolve it, the model will be available on HF.
\n
FINAL UPDATE…
\nI tested something more in depth. The problem is, I can’t upload files LARGER than 10 Megabytes!
\nI used THIS python script to create dummy files:
\nimport os
\nos.chdir(os.path.dirname(os.path.abspath(file)))
#zeros = 524259 # 900 MB
\n#zeros = 524317 # also 900 MB
\ncomment=“”""
\nx = 1800
\nfile_size = zeros * x
\nfile_name = 0.5 * x
\nwith open(f""{str(file_name)} mb.txt"", “w”) as f:
\nf.write(“0” * file_size)
\nx = x + 1
\n“”""
\n#print(f""zeros = {round((524259+524317)/2)}"")
zeros = 524288
\nx = 1
\nwhile(x < 201):
\nfile_size = zeros * x
\nfile_name = 0.5 * x
\nwith open(f""{str(file_name)} mb.txt"", “w”) as f:
\nf.write(“0” * file_size)
\nx = x + 1
print(“Files created: (size) mb.txt (0.5 MB of zeros incrementals)”)
\nthe 10.5 MB file BROKE it, but the 10 MB file WORKED!
\nTHAT MEANS THE PROBLEM IS DIRECTLY ON THEIR END, SOME PIECE OF CODE SAYS:
\nif(filesize > 10 MB):
\ndo something
\nelse:
\ndo something different
It’s NOT my computer, it’s some glitch in THEIR system. something above 10 MB breaks it for some reason!
\nOh well, I use git on Google Colab anyway. No big deal I guess…
\nMy proof:
\n\nI also found documentation here:
\n\n\nGit LFS automatically handles files larger than 10MB. But for very large files (>5GB), you need to install a custom transfer agent for Git LFS:
\nCopied
\nhuggingface-cli lfs-enable-largefiles
\nYou should install this for each repository that has a very large file. Once installed, you’ll be able to push files larger than 5GB.
\nThe commit context manager handles four of the most common Git commands: pull, add, commit, and push. git-lfs automatically tracks any file larger than 10MB. In the following example, the commit context manager:
That SPECIFIC number is mentioned here.
', 'post_number': 18, 'post_type': 1, 'posts_count': 20, 'updated_at': '2025-05-20T20:59:23.690Z', 'reply_count': 0, 'reply_to_post_number': 17, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 25.6, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'Phoenix Storm Jr.', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/huggingface_hub/v0.17.1/en/guides/upload#hub-repository-size-limitations', 'internal': False, 'reflection': False, 'title': 'Upload files to the Hub', 'clicks': 1}, {'url': 'https://huggingface.co/PhoenixStormJr/test-upload-length/tree/main', 'internal': False, 'reflection': False, 'title': 'PhoenixStormJr/test-upload-length at main', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64378, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/18', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 223239, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-21T05:17:32.005Z', 'cooked': 'Hmm… It seems to be a bug on the Hub side related to LFS…
In a Windows environment, the explanation is simple: you need to install LFS and git itself using the installer, but I don’t think that’s the case here.
\n\n', 'post_number': 19, 'post_type': 1, 'posts_count': 20, 'updated_at': '2025-05-21T05:17:32.005Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://git-scm.com/downloads/win', 'internal': False, 'reflection': False, 'title': 'Git - Downloading Package', 'clicks': 0}, {'url': 'https://git-lfs.com/', 'internal': False, 'reflection': False, 'title': 'Git Large File Storage | Git Large File Storage (LFS) replaces large files such as audio samples, videos, datasets, and graphics with text pointers inside Git, while storing the file contents on a remote server like GitHub.com or GitHub Enterprise.', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/19', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 223604, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-23T00:14:02.304Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 20, 'post_type': 3, 'posts_count': 20, 'updated_at': '2025-05-23T00:14:02.304Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 106539, 'topic_slug': 'cant-upload-my-model-stuck-on-hashing', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/cant-upload-my-model-stuck-on-hashing/106539/20', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","The title says pretty much everything. I was able to upload with a Google Colab hack, but normally, I can’t. I attached the files down below. Can anyone figure out what the deal is?
+I “fixed” the problem by uploading them with google colab, but I don’t like this solution. Why won’t it upload normally? Here is the colab link:
+ +Here is the screenshot showing the huggingface refusing to hash:
+And here are the files that wouldn’t hash:
+ + +What’s going on?
","Hmm… It seems to be a bug on the Hub side related to LFS…
In a Windows environment, the explanation is simple: you need to install LFS and git itself using the installer, but I don’t think that’s the case here.
+ +" +How to organize hundreds of pre-trained models,https://discuss.huggingface.co/t/how-to-organize-hundreds-of-pre-trained-models/42682,42682,5,2023-06-09 16:37:47.869000+00:00,"[{'id': 73328, 'name': 'Adam Stewart', 'username': 'ajstewart', 'avatar_template': '/user_avatar/discuss.huggingface.co/ajstewart/{size}/47937_2.png', 'created_at': '2023-06-09T16:37:47.925Z', 'cooked': 'We (torchgeo (TorchGeo)) are working on a project that will generate 100+ pre-trained models. In the past, we’ve made a separate repository for each model, but with 100+ models we’ve started to wonder whether or not it would make more sense to stuff all of our models in a few repos instead of having 100+ separate repos. What features or functionality would we lose by doing so? Our users primarily load weights through the TorchGeo library (using timm or smp) and don’t even know that HF exists, it’s just the place we chose to distribute the files.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2023-06-09T16:37:47.925Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 60, 'reads': 12, 'readers_count': 11, 'score': 332.4, 'yours': False, 'topic_id': 42682, 'topic_slug': 'how-to-organize-hundreds-of-pre-trained-models', 'display_username': 'Adam Stewart', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://hf.co/torchgeo', 'internal': False, 'reflection': False, 'title': 'torchgeo (TorchGeo)', 'clicks': 2}, {'url': 'https://discuss.huggingface.co/t/how-to-handle-very-large-datasets/42686', 'internal': True, 'reflection': True, 'title': 'How to handle very large datasets', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 21698, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-organize-hundreds-of-pre-trained-models/42682/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 223270, 'name': 'Lucain Pouget', 'username': 'Wauplin', 'avatar_template': '/user_avatar/discuss.huggingface.co/wauplin/{size}/40815_2.png', 'created_at': '2025-05-21T07:21:38.516Z', 'cooked': 'Late to the party, but it’s always recommended to do 1 pretrained model == 1 repo. It allows to have a download counter per model (allowing you to know which models are getting more traction), better discoverability for users on the Hub, dedicated community tabs per variant, etc.
\n(related: Add TorchGeo to libraries by isaaccorley · Pull Request #1464 · huggingface/huggingface.js · GitHub)
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-21T07:21:38.516Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 6, 'readers_count': 5, 'score': 36.2, 'yours': False, 'topic_id': 42682, 'topic_slug': 'how-to-organize-hundreds-of-pre-trained-models', 'display_username': 'Lucain Pouget', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/huggingface.js/pull/1464#discussion_r2098481444', 'internal': False, 'reflection': False, 'title': 'Add TorchGeo to libraries by isaaccorley · Pull Request #1464 · huggingface/huggingface.js · GitHub', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 9207, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-organize-hundreds-of-pre-trained-models/42682/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 223372, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-21T19:21:51.055Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-21T19:21:51.055Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 42682, 'topic_slug': 'how-to-organize-hundreds-of-pre-trained-models', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-organize-hundreds-of-pre-trained-models/42682/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","We (torchgeo (TorchGeo)) are working on a project that will generate 100+ pre-trained models. In the past, we’ve made a separate repository for each model, but with 100+ models we’ve started to wonder whether or not it would make more sense to stuff all of our models in a few repos instead of having 100+ separate repos. What features or functionality would we lose by doing so? Our users primarily load weights through the TorchGeo library (using timm or smp) and don’t even know that HF exists, it’s just the place we chose to distribute the files.
","Late to the party, but it’s always recommended to do 1 pretrained model == 1 repo. It allows to have a download counter per model (allowing you to know which models are getting more traction), better discoverability for users on the Hub, dedicated community tabs per variant, etc.
+(related: Add TorchGeo to libraries by isaaccorley · Pull Request #1464 · huggingface/huggingface.js · GitHub)
" +How to iterate over values of a column in the IterableDataset?,https://discuss.huggingface.co/t/how-to-iterate-over-values-of-a-column-in-the-iterabledataset/135649,135649,10,2025-01-14 11:33:40.731000+00:00,"[{'id': 195452, 'name': 'Svyatoslav V. Pchelintsev', 'username': 'Innovator2K', 'avatar_template': '/user_avatar/discuss.huggingface.co/innovator2k/{size}/38148_2.png', 'created_at': '2025-01-14T11:33:40.784Z', 'cooked': 'Suppose we have a simple iterable dataset from the documentation:
\ndef gen():\n yield {""text"": ""Good"", ""label"": 0}\n yield {""text"": ""Bad"", ""label"": 1}\n\nds = IterableDataset.from_generator(gen)\nand suppose I want to iterate over the ""text"" column values. An obvious solution can be the following:
column_values_only_ds = map(lambda x: x[""text""], ds)\nBut the problem with this solution is that map is not an iterable, i.e., it cannot be re-iterated:
for v in column_values_only_ds:\n print(v) # Prints ""Good"" and ""Bad""\nfor v in column_values_only_ds:\n print(v) # Prints nothing\nSo, how can I create an iterable that returns only column values?
\nP.S. I’m building a single interface for running experiments with different models and, e.g., FastText requires only lists of strings, not dictionaries.
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-01-14T11:33:40.784Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 74, 'reads': 10, 'readers_count': 9, 'score': 367.0, 'yours': False, 'topic_id': 135649, 'topic_slug': 'how-to-iterate-over-values-of-a-column-in-the-iterabledataset', 'display_username': 'Svyatoslav V. Pchelintsev', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.IterableDataset.from_generator', 'internal': False, 'reflection': False, 'title': 'Main classes', 'clicks': 2}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 35404, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-iterate-over-values-of-a-column-in-the-iterabledataset/135649/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 195465, 'name': 'Alan turner', 'username': 'Alanturner2', 'avatar_template': '/user_avatar/discuss.huggingface.co/alanturner2/{size}/37542_2.png', 'created_at': '2025-01-14T13:10:11.600Z', 'cooked': 'Hi there!
If you want to iterate over just the ""text"" column in your IterableDataset and make sure it can be re-iterated (unlike map), you can use a generator function. This way, you’ll always get a fresh iterable whenever you need it.
Here’s how you can do it:
\nfrom datasets import IterableDataset\n\n# Your original dataset generator\ndef gen():\n yield {""text"": ""Good"", ""label"": 0}\n yield {""text"": ""Bad"", ""label"": 1}\n\nds = IterableDataset.from_generator(gen)\n\n# A function to pull only the ""text"" values\ndef extract_text_column(dataset):\n for item in dataset:\n yield item[""text""]\n\n# A callable that gives you a fresh iterator each time\ncolumn_values_only_ds = lambda: extract_text_column(ds)\n\n# Now, let\'s iterate over the ""text"" column\nfor v in column_values_only_ds():\n print(v) # Prints ""Good"" and ""Bad""\n\n# You can do it again without issues!\nfor v in column_values_only_ds():\n print(v) # Prints ""Good"" and ""Bad"" again\nextract_text_column(dataset) is like a recipe to grab just the ""text"" values one at a time.column_values_only_ds(), it gives you a brand-new iterator. So, no matter how many times you loop, it works!I hope this clears things up and helps you with your project. Feel free to reach out if you have more questions. Happy coding!
Thank you for the answer!
\nWhile this works, it loses the functionality of the IterableDataset (its methods and attributes are no longer accessible), so I hoped for a built in Datasets solution, but your answer suggests that there is no such functionality. OK.
By the way, something like this should also work:
\nclass IterableDatasetColumnGetter:\n def __init__(self, dataset: IterableDataset, column_name: str) -> None:\n self.dataset = dataset\n self.column_name = column_name\n\n def __iter__(self) -> Iterator:\n return iter(map(lambda x: x[self.column_name], self.dataset))\n\niterable_column_values_only_ds = IterableDatasetColumnGetter(ds, ""text"")\n\nfor v in iterable_column_values_only_ds:\n print(v) # Prints ""Good"" and ""Bad""\n\nfor v in iterable_column_values_only_ds:\n print(v) # Prints ""Good"" and ""Bad"" again\nbut again it looks like it is not a good solution due to the loss of the original functionality.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-01-14T14:11:01.305Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 10, 'readers_count': 9, 'score': 42.0, 'yours': False, 'topic_id': 135649, 'topic_slug': 'how-to-iterate-over-values-of-a-column-in-the-iterabledataset', 'display_username': 'Svyatoslav V. Pchelintsev', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76958, 'username': 'Alanturner2', 'name': 'Alan turner', 'avatar_template': '/user_avatar/discuss.huggingface.co/alanturner2/{size}/37542_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 35404, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-iterate-over-values-of-a-column-in-the-iterabledataset/135649/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 195574, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-01-15T02:07:22.561Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-01-15T02:07:22.561Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 8, 'readers_count': 7, 'score': 6.6, 'yours': False, 'topic_id': 135649, 'topic_slug': 'how-to-iterate-over-values-of-a-column-in-the-iterabledataset', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-iterate-over-values-of-a-column-in-the-iterabledataset/135649/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 198129, 'name': 'Quentin Lhoest', 'username': 'lhoestq', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png', 'created_at': '2025-01-27T10:42:47.008Z', 'cooked': 'Hi ! Could it be interesting to implement a IterableColumn ? What do you think of something like this ?
\ndef gen():\n yield {""text"": ""Good"", ""label"": 0}\n yield {""text"": ""Bad"", ""label"": 1}\n\nds = IterableDataset.from_generator(gen)\ntexts = ds[""text""] # `texts` is an IterableColumn object\n\nfor v in texts:\n print(v)\nIf you like this API, feel free to suggest it in an issue on gtihub or open a PR
Hi ! it’s now possible to iterate on a column directly, thanks @Innovator2K !
\nThe PR is here Implementation of iteration over values of a column in an IterableDataset object by TopCoder2K · Pull Request #7564 · huggingface/datasets · GitHub and the feature will be available in the next release
>>> from datasets import load_dataset\n>>> dataset = load_dataset(""allenai/c4"", ""en"", streaming=True, split=""train"")\n>>> print(next(iter(dataset[""text""])))\nBeginners BBQ Class Taking Place in Missoula!...\nSuppose we have a simple iterable dataset from the documentation:
+def gen():
+ yield {""text"": ""Good"", ""label"": 0}
+ yield {""text"": ""Bad"", ""label"": 1}
+
+ds = IterableDataset.from_generator(gen)
+and suppose I want to iterate over the ""text"" column values. An obvious solution can be the following:
column_values_only_ds = map(lambda x: x[""text""], ds)
+But the problem with this solution is that map is not an iterable, i.e., it cannot be re-iterated:
for v in column_values_only_ds:
+ print(v) # Prints ""Good"" and ""Bad""
+for v in column_values_only_ds:
+ print(v) # Prints nothing
+So, how can I create an iterable that returns only column values?
+P.S. I’m building a single interface for running experiments with different models and, e.g., FastText requires only lists of strings, not dictionaries.
","Hi there!
If you want to iterate over just the ""text"" column in your IterableDataset and make sure it can be re-iterated (unlike map), you can use a generator function. This way, you’ll always get a fresh iterable whenever you need it.
Here’s how you can do it:
+from datasets import IterableDataset
+
+# Your original dataset generator
+def gen():
+ yield {""text"": ""Good"", ""label"": 0}
+ yield {""text"": ""Bad"", ""label"": 1}
+
+ds = IterableDataset.from_generator(gen)
+
+# A function to pull only the ""text"" values
+def extract_text_column(dataset):
+ for item in dataset:
+ yield item[""text""]
+
+# A callable that gives you a fresh iterator each time
+column_values_only_ds = lambda: extract_text_column(ds)
+
+# Now, let's iterate over the ""text"" column
+for v in column_values_only_ds():
+ print(v) # Prints ""Good"" and ""Bad""
+
+# You can do it again without issues!
+for v in column_values_only_ds():
+ print(v) # Prints ""Good"" and ""Bad"" again
+extract_text_column(dataset) is like a recipe to grab just the ""text"" values one at a time.column_values_only_ds(), it gives you a brand-new iterator. So, no matter how many times you loop, it works!I hope this clears things up and helps you with your project. Feel free to reach out if you have more questions. Happy coding!
Hi,
\nI’m quite familiar with the Huggingface ecosystem and I used it a lot.
\nHowever, I cannot find resources/models / tutorials for coreference resolution except for neuralcoref which last commit was years ago…
\nI also saw some models but there is not any clue on how to use them (I guess a TokenClassification Head ?)
\nDoes anyone have any starting point for implementing a coreference resolution pipeline?
\n(I will start will neuralcoref if there is nothing better)
Thanks in advance for any help,
\nHave a great day.
Hi,
\nI suggest to take a look at this repo: GitHub - mandarjoshi90/coref: BERT for Coreference Resolution
\nIt includes multiple models (BERT, SpanBERT) fine-tuned on OntoNotes, an important benchmark for coreference resolution.
\nThere’s also a demo notebook, showcasing how to run inference for a new piece of text to find all entity clusters.
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2021-11-08T08:36:40.298Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 163, 'reads': 53, 'readers_count': 52, 'score': 875.6, 'yours': False, 'topic_id': 11394, 'topic_slug': 'coreference-resolution', 'display_username': 'Niels Rogge', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/mandarjoshi90/coref', 'internal': False, 'reflection': False, 'title': 'GitHub - mandarjoshi90/coref: BERT for Coreference Resolution', 'clicks': 632}, {'url': 'https://colab.research.google.com/drive/1SlERO9Uc9541qv6yH26LJz5IM9j7YVra#scrollTo=H0xPknceFORt', 'internal': False, 'reflection': False, 'title': 'Google Colab', 'clicks': 314}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 205, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/coreference-resolution/11394/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}, {'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 222878, 'name': 'Anushka', 'username': 'anuyash49', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/4af34b/{size}.png', 'created_at': '2025-05-19T06:05:54.578Z', 'cooked': 'not updated. can’t run SpanBERT
', 'post_number': 3, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-19T06:05:54.578Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 3, 'readers_count': 2, 'score': 45.6, 'yours': False, 'topic_id': 11394, 'topic_slug': 'coreference-resolution', 'display_username': 'Anushka', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 205, 'username': 'nielsr', 'name': 'Niels Rogge', 'avatar_template': '/user_avatar/discuss.huggingface.co/nielsr/{size}/39617_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94410, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/coreference-resolution/11394/3', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I’m quite familiar with the Huggingface ecosystem and I used it a lot.
+However, I cannot find resources/models / tutorials for coreference resolution except for neuralcoref which last commit was years ago…
+I also saw some models but there is not any clue on how to use them (I guess a TokenClassification Head ?)
+Does anyone have any starting point for implementing a coreference resolution pipeline?
+(I will start will neuralcoref if there is nothing better)
Thanks in advance for any help,
+Have a great day.
Hi,
+I suggest to take a look at this repo: GitHub - mandarjoshi90/coref: BERT for Coreference Resolution
+It includes multiple models (BERT, SpanBERT) fine-tuned on OntoNotes, an important benchmark for coreference resolution.
+There’s also a demo notebook, showcasing how to run inference for a new piece of text to find all entity clusters.
" +Best model to extract text from old Church records written in cursive?,https://discuss.huggingface.co/t/best-model-to-extract-text-from-old-church-records-written-in-cursive/155677,155677,13,2025-05-17 18:07:35.911000+00:00,"[{'id': 222667, 'name': 'Danijel Meglen', 'username': 'podtalnica', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/p/65b543/{size}.png', 'created_at': '2025-05-17T18:07:35.963Z', 'cooked': 'Hello! I have a bunch of Church records that I got from Matricula Online (a website that stores church registers like books of birth, marriage and death). They are from 16th all the way to early 20th century. I would like to store their contents in a .txt file. Records are written in cursive in a mix between Slovene and German. Here’s a random page so you can see what I mean. I have a GTX 1060 6GB so naturally I would like a model that I can run on my computer without major performance issues. What would be the best model to do this? Thank you in advance!
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-17T18:07:35.963Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 34, 'reads': 7, 'readers_count': 6, 'score': 171.4, 'yours': False, 'topic_id': 155677, 'topic_slug': 'best-model-to-extract-text-from-old-church-records-written-in-cursive', 'display_username': 'Danijel Meglen', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://data.matricula-online.eu/en/slovenia/ljubljana/zagradec/04415/?pg=12', 'internal': False, 'reflection': False, 'title': 'Krstna knjiga / Taufbuch - 04415 | Zagradec | Nadškofijski arhiv Ljubljana | Slovenia | Matricula Online', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 94287, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/best-model-to-extract-text-from-old-church-records-written-in-cursive/155677/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 222716, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-18T00:16:26.225Z', 'cooked': 'Basically, this task can be performed using VLM, but recognizing actual handwritten characters and text is quite difficult. I recommend trying out various models online and using the ones that work well locally. With VRAM savings through quantization, there are models that can run with 6GB.
\n\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-18T00:16:26.225Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 7, 'readers_count': 6, 'score': 21.4, 'yours': False, 'topic_id': 155677, 'topic_slug': 'best-model-to-extract-text-from-old-church-records-written-in-cursive', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/microsoft/trocr-large-handwritten', 'internal': False, 'reflection': False, 'title': 'microsoft/trocr-large-handwritten · Hugging Face', 'clicks': 5}, {'url': 'https://huggingface.co/spaces?sort=trending&search=vl', 'internal': False, 'reflection': False, 'title': 'Spaces - Hugging Face', 'clicks': 1}, {'url': 'https://discuss.huggingface.co/t/handwriting-recognition-cant-recognize-multiline-words/39422', 'internal': True, 'reflection': False, 'title': ""Handwriting recognition. Can't recognize multiline words"", 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/handwritten-ocr-w-confidence-scores/143476', 'internal': True, 'reflection': False, 'title': 'Handwritten OCR w/ confidence scores', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/best-model-to-extract-text-from-old-church-records-written-in-cursive/155677/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 222778, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-18T12:17:19.657Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-18T12:17:19.657Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 5, 'readers_count': 4, 'score': 6.0, 'yours': False, 'topic_id': 155677, 'topic_slug': 'best-model-to-extract-text-from-old-church-records-written-in-cursive', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/best-model-to-extract-text-from-old-church-records-written-in-cursive/155677/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello! I have a bunch of Church records that I got from Matricula Online (a website that stores church registers like books of birth, marriage and death). They are from 16th all the way to early 20th century. I would like to store their contents in a .txt file. Records are written in cursive in a mix between Slovene and German. Here’s a random page so you can see what I mean. I have a GTX 1060 6GB so naturally I would like a model that I can run on my computer without major performance issues. What would be the best model to do this? Thank you in advance!
","Basically, this task can be performed using VLM, but recognizing actual handwritten characters and text is quite difficult. I recommend trying out various models online and using the ones that work well locally. With VRAM savings through quantization, there are models that can run with 6GB.
+ + + +" +Can I write to the file system?,https://discuss.huggingface.co/t/can-i-write-to-the-file-system/155246,155246,24,2025-05-14 21:45:09.585000+00:00,"[{'id': 222086, 'name': 'Pablo Villanueva Domingo', 'username': 'PabloVD', 'avatar_template': '/user_avatar/discuss.huggingface.co/pablovd/{size}/34178_2.png', 'created_at': '2025-05-14T21:45:09.637Z', 'cooked': 'I have an app where I need to write files to the file system, like:
\nos.makedirs(work_dir)\nIs that possible? I tried with a docker image but I got a PermissionError: [Errno 13] Permission denied in that line. Any way to overcome that?
I think you can basically access the directory under /home/user/ (or possibly /home/ ?) using that method. There is no way to access a path higher up…
(This also causes an error in Dockerfile’s WORKDIR, etc.)
That was the reason! I needed to create an user and work in the user folder. The steps to follow are explained here.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-16T08:36:31.656Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 155246, 'topic_slug': 'can-i-write-to-the-file-system', 'display_username': 'Pablo Villanueva Domingo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/hub/spaces-sdks-docker', 'internal': False, 'reflection': False, 'title': 'Docker Spaces', 'clicks': 8}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 69899, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-write-to-the-file-system/155246/3', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 222553, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-16T20:36:50.624Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-05-16T20:36:50.624Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 2, 'readers_count': 1, 'score': 10.4, 'yours': False, 'topic_id': 155246, 'topic_slug': 'can-i-write-to-the-file-system', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/can-i-write-to-the-file-system/155246/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have an app where I need to write files to the file system, like:
+os.makedirs(work_dir)
+Is that possible? I tried with a docker image but I got a PermissionError: [Errno 13] Permission denied in that line. Any way to overcome that?
I think you can basically access the directory under /home/user/ (or possibly /home/ ?) using that method. There is no way to access a path higher up…
(This also causes an error in Dockerfile’s WORKDIR, etc.)
Hi,
\nI just finetuned Tiny-Llama as tiny-sajar, a little experiment to test finetuning. Running the following code in google colab:
from transformers import AutoModelForCausalLM, AutoTokenizer\n\n# Replace with your model\'s path on the Hub\nmodel = AutoModelForCausalLM.from_pretrained(""Dagriffpatchfan/tiny-sajar"")\ntokenizer = AutoTokenizer.from_pretrained(""Dagriffpatchfan/tiny-sajar"")\n\nWorked perfectly, loading the model. I was then able to run the following code:
\nquestions = [\n ""Questions here"",\n]\n\nfor question in questions:\n prompt = f""{question}""\n inputs = tokenizer(prompt, return_tensors=""pt"")\n outputs = model.generate(\n inputs.input_ids,\n max_length=100, # Maximum number of tokens to generate\n num_return_sequences=1, # Number of separate completions to generate\n temperature=0.7, # Sampling temperature (lower is more focused, higher is more random)\n top_p=0.9, # Nucleus sampling\n do_sample=True # Enable sampling\n )\n\n # Decode the generated text\n generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)\n print(f""**{question}**\\n{generated_text}\\n"")\n\nWhich generated text as expected. I went to try this in a jupyterlab space and to my complete surprise I got the following error when I tried to load the model:
\n--------------------------------------------------------------------------- ValueError Traceback (most recent call last) Cell In[7], line 4 1 from transformers import AutoModelForCausalLM, AutoTokenizer 3 # Replace with your model’s path on the Hub ----> 4 model = AutoModelForCausalLM.from_pretrained(“Dagriffpatchfan/tiny-sajar”) 5 tokenizer = AutoTokenizer.from_pretrained(“Dagriffpatchfan/tiny-sajar”) 7 questions = [ 8 “Who are you, and what is your role in the story?”, 9 “How did you come to know David and the Avengers?”, (…) 17 “If you had to pick one person to go on a mission with, who would it be and why?” 18 ] File ~/miniconda/lib/python3.9/site-packages/transformers/models/auto/auto_factory.py:531, in _BaseAutoModelClass.from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs) 528 if kwargs.get(“quantization_config”, None) is not None: 529 _ = kwargs.pop(“quantization_config”) → 531 config, kwargs = AutoConfig.from_pretrained( 532 pretrained_model_name_or_path, 533 return_unused_kwargs=True, 534 trust_remote_code=trust_remote_code, 535 code_revision=code_revision, 536 _commit_hash=commit_hash, 537 **hub_kwargs, 538 **kwargs, 539 ) 541 # if torch_dtype=auto was passed here, ensure to pass it on 542 if kwargs_orig.get(“torch_dtype”, None) == “auto”: File ~/miniconda/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py:1151, in AutoConfig.from_pretrained(cls, pretrained_model_name_or_path, **kwargs) 1148 if pattern in str(pretrained_model_name_or_path): 1149 return CONFIG_MAPPING[pattern].from_dict(config_dict, **unused_kwargs) → 1151 raise ValueError( 1152 f""Unrecognized model in {pretrained_model_name_or_path}. "" 1153 f""Should have a model_type key in its {CONFIG_NAME}, or contain one of the following strings "" 1154 f""in its name: {\', \'.join(CONFIG_MAPPING.keys())}"" 1155 ) ValueError: Unrecognized model in Dagriffpatchfan/tiny-sajar. Should have a model_type key in its config.json, or contain one of the following strings in its name: albert, align, altclip, aria, aria_text, audio-spectrogram-transformer, autoformer, aya_vision, bamba, bark, bart, beit, bert, bert-generation, big_bird, bigbird_pegasus, biogpt, bit, blenderbot, blenderbot-small, blip, blip-2, bloom, bridgetower, bros, camembert, canine, chameleon, chinese_clip, chinese_clip_vision_model, clap, clip, clip_text_model, clip_vision_model, clipseg, clvp, code_llama, codegen, cohere, cohere2, colpali, conditional_detr, convbert, convnext, convnextv2, cpmant, ctrl, cvt, dab-detr, dac, data2vec-audio, data2vec-text, data2vec-vision, dbrx, deberta, deberta-v2, decision_transformer, deepseek_v3, deformable_detr, deit, depth_anything, depth_pro, deta, detr, diffllama, dinat, dinov2, dinov2_with_registers, distilbert, donut-swin, dpr, dpt, efficientformer, efficientnet, electra, emu3, encodec, encoder-decoder, ernie, ernie_m, esm, falcon, falcon_mamba, fastspeech2_conformer, flaubert, flava, fnet, focalnet, fsmt, funnel, fuyu, gemma, gemma2, gemma3, gemma3_text, git, glm, glm4, glpn, got_ocr2, gpt-sw3, gpt2, gpt_bigcode, gpt_neo, gpt_neox, gpt_neox_japanese, gptj, gptsan-japanese, granite, granitemoe, granitemoeshared, granitevision, graphormer, grounding-dino, groupvit, helium, hiera, hubert, ibert, idefics, idefics2, idefics3, idefics3_vision, ijepa, imagegpt, informer, instructblip, instructblipvideo, jamba, jetmoe, jukebox, kosmos-2, layoutlm, layoutlmv2, layoutlmv3, led, levit, lilt, llama, llama4, llama4_text, llava, llava_next, llava_next_video, llava_onevision, longformer, longt5, luke, lxmert, m2m_100, mamba, mamba2, marian, markuplm, mask2former, maskformer, maskformer-swin, mbart, mctct, mega, megatron-bert, mgp-str, mimi, mistral, mistral3, mixtral, mllama, mobilebert, mobilenet_v1, mobilenet_v2, mobilevit, mobilevitv2, modernbert, moonshine, moshi, mpnet, mpt, mra, mt5, musicgen, musicgen_melody, mvp, nat, nemotron, nezha, nllb-moe, nougat, nystromformer, olmo, olmo2, olmoe, omdet-turbo, oneformer, open-llama, openai-gpt, opt, owlv2, owlvit, paligemma, patchtsmixer, patchtst, pegasus, pegasus_x, perceiver, persimmon, phi, phi3, phi4_multimodal, phimoe, pix2struct, pixtral, plbart, poolformer, pop2piano, prompt_depth_anything, prophetnet, pvt, pvt_v2, qdqbert, qwen2, qwen2_5_vl, qwen2_audio, qwen2_audio_encoder, qwen2_moe, qwen2_vl, qwen3, qwen3_moe, rag, realm, recurrent_gemma, reformer, regnet, rembert, resnet, retribert, roberta, roberta-prelayernorm, roc_bert, roformer, rt_detr, rt_detr_resnet, rt_detr_v2, rwkv, sam, sam_vision_model, seamless_m4t, seamless_m4t_v2, segformer, seggpt, sew, sew-d, shieldgemma2, siglip, siglip2, siglip_vision_model, smolvlm, smolvlm_vision, speech-encoder-decoder, speech_to_text, speech_to_text_2, speecht5, splinter, squeezebert, stablelm, starcoder2, superglue, superpoint, swiftformer, swin, swin2sr, swinv2, switch_transformers, t5, table-transformer, tapas, textnet, time_series_transformer, timesformer, timm_backbone, timm_wrapper, trajectory_transformer, transfo-xl, trocr, tvlt, tvp, udop, umt5, unispeech, unispeech-sat, univnet, upernet, van, video_llava, videomae, vilt, vipllava, vision-encoder-decoder, vision-text-dual-encoder, visual_bert, vit, vit_hybrid, vit_mae, vit_msn, vitdet, vitmatte, vitpose, vitpose_backbone, vits, vivit, wav2vec2, wav2vec2-bert, wav2vec2-conformer, wavlm, whisper, xclip, xglm, xlm, xlm-prophetnet, xlm-roberta, xlm-roberta-xl, xlnet, xmod, yolos, yoso, zamba, zamba2, zoedepth
I found this very confusing…does anyone know what I am experiencing?
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-08T08:37:41.764Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 32, 'reads': 4, 'readers_count': 3, 'score': 155.8, 'yours': False, 'topic_id': 154082, 'topic_slug': 'model-loading-in-colab-but-not-jupyterlab', 'display_username': 'David Mathew', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://dagriffpatchfan-jupyterlab.hf.space/lab/tree/~/miniconda/lib/python3.9/site-packages/transformers/models/auto/auto_factory.py#line=530', 'internal': False, 'reflection': False, 'clicks': 0}, {'url': 'https://dagriffpatchfan-jupyterlab.hf.space/lab/tree/~/miniconda/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py#line=1150', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90119, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-loading-in-colab-but-not-jupyterlab/154082/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 220688, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-08T23:55:50.918Z', 'cooked': 'Since it includes models close to the latest ones such as Gemma 3, the Transoformers version is likely to be almost the latest. In fact, even older Transoformers models should work with the Llama architecture. This is indeed a strange error. The cause is probably not the code or the model itself.
\nThere seems to be a possibility of errors occurring in hf_transfer related to Jupyter. In other words, there may be an error in the download.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-08T23:55:50.918Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 5.8, 'yours': False, 'topic_id': 154082, 'topic_slug': 'model-loading-in-colab-but-not-jupyterlab', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/autotokenizer-from-pretrained-suddenly-raises-an-error/153809/4', 'internal': True, 'reflection': False, 'title': 'AutoTokenizer.from_pretrained() suddenly raises an error', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/model-loading-in-colab-but-not-jupyterlab/154082/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221277, 'name': 'David Mathew', 'username': 'Dagriffpatchfan', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/d07c76/{size}.png', 'created_at': '2025-05-11T22:21:32.620Z', 'cooked': 'So I should set
\nexport HF_HUB_ENABLE_HF_TRANSFER=1
\nto 0 instead of 1?
Yea. Or maybe try reinstalling hf_transfer. If that’s the cause.
pip install -U hf_transfer hf_xet\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-15T23:33:42.138Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 154082, 'topic_slug': 'model-loading-in-colab-but-not-jupyterlab', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/model-loading-in-colab-but-not-jupyterlab/154082/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I just finetuned Tiny-Llama as tiny-sajar, a little experiment to test finetuning. Running the following code in google colab:
from transformers import AutoModelForCausalLM, AutoTokenizer
+
+# Replace with your model's path on the Hub
+model = AutoModelForCausalLM.from_pretrained(""Dagriffpatchfan/tiny-sajar"")
+tokenizer = AutoTokenizer.from_pretrained(""Dagriffpatchfan/tiny-sajar"")
+
+Worked perfectly, loading the model. I was then able to run the following code:
+questions = [
+ ""Questions here"",
+]
+
+for question in questions:
+ prompt = f""{question}""
+ inputs = tokenizer(prompt, return_tensors=""pt"")
+ outputs = model.generate(
+ inputs.input_ids,
+ max_length=100, # Maximum number of tokens to generate
+ num_return_sequences=1, # Number of separate completions to generate
+ temperature=0.7, # Sampling temperature (lower is more focused, higher is more random)
+ top_p=0.9, # Nucleus sampling
+ do_sample=True # Enable sampling
+ )
+
+ # Decode the generated text
+ generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+ print(f""**{question}**\n{generated_text}\n"")
+
+Which generated text as expected. I went to try this in a jupyterlab space and to my complete surprise I got the following error when I tried to load the model:
+--------------------------------------------------------------------------- ValueError Traceback (most recent call last) Cell In[7], line 4 1 from transformers import AutoModelForCausalLM, AutoTokenizer 3 # Replace with your model’s path on the Hub ----> 4 model = AutoModelForCausalLM.from_pretrained(“Dagriffpatchfan/tiny-sajar”) 5 tokenizer = AutoTokenizer.from_pretrained(“Dagriffpatchfan/tiny-sajar”) 7 questions = [ 8 “Who are you, and what is your role in the story?”, 9 “How did you come to know David and the Avengers?”, (…) 17 “If you had to pick one person to go on a mission with, who would it be and why?” 18 ] File ~/miniconda/lib/python3.9/site-packages/transformers/models/auto/auto_factory.py:531, in _BaseAutoModelClass.from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs) 528 if kwargs.get(“quantization_config”, None) is not None: 529 _ = kwargs.pop(“quantization_config”) → 531 config, kwargs = AutoConfig.from_pretrained( 532 pretrained_model_name_or_path, 533 return_unused_kwargs=True, 534 trust_remote_code=trust_remote_code, 535 code_revision=code_revision, 536 _commit_hash=commit_hash, 537 **hub_kwargs, 538 **kwargs, 539 ) 541 # if torch_dtype=auto was passed here, ensure to pass it on 542 if kwargs_orig.get(“torch_dtype”, None) == “auto”: File ~/miniconda/lib/python3.9/site-packages/transformers/models/auto/configuration_auto.py:1151, in AutoConfig.from_pretrained(cls, pretrained_model_name_or_path, **kwargs) 1148 if pattern in str(pretrained_model_name_or_path): 1149 return CONFIG_MAPPING[pattern].from_dict(config_dict, **unused_kwargs) → 1151 raise ValueError( 1152 f""Unrecognized model in {pretrained_model_name_or_path}. "" 1153 f""Should have a model_type key in its {CONFIG_NAME}, or contain one of the following strings "" 1154 f""in its name: {', '.join(CONFIG_MAPPING.keys())}"" 1155 ) ValueError: Unrecognized model in Dagriffpatchfan/tiny-sajar. Should have a model_type key in its config.json, or contain one of the following strings in its name: albert, align, altclip, aria, aria_text, audio-spectrogram-transformer, autoformer, aya_vision, bamba, bark, bart, beit, bert, bert-generation, big_bird, bigbird_pegasus, biogpt, bit, blenderbot, blenderbot-small, blip, blip-2, bloom, bridgetower, bros, camembert, canine, chameleon, chinese_clip, chinese_clip_vision_model, clap, clip, clip_text_model, clip_vision_model, clipseg, clvp, code_llama, codegen, cohere, cohere2, colpali, conditional_detr, convbert, convnext, convnextv2, cpmant, ctrl, cvt, dab-detr, dac, data2vec-audio, data2vec-text, data2vec-vision, dbrx, deberta, deberta-v2, decision_transformer, deepseek_v3, deformable_detr, deit, depth_anything, depth_pro, deta, detr, diffllama, dinat, dinov2, dinov2_with_registers, distilbert, donut-swin, dpr, dpt, efficientformer, efficientnet, electra, emu3, encodec, encoder-decoder, ernie, ernie_m, esm, falcon, falcon_mamba, fastspeech2_conformer, flaubert, flava, fnet, focalnet, fsmt, funnel, fuyu, gemma, gemma2, gemma3, gemma3_text, git, glm, glm4, glpn, got_ocr2, gpt-sw3, gpt2, gpt_bigcode, gpt_neo, gpt_neox, gpt_neox_japanese, gptj, gptsan-japanese, granite, granitemoe, granitemoeshared, granitevision, graphormer, grounding-dino, groupvit, helium, hiera, hubert, ibert, idefics, idefics2, idefics3, idefics3_vision, ijepa, imagegpt, informer, instructblip, instructblipvideo, jamba, jetmoe, jukebox, kosmos-2, layoutlm, layoutlmv2, layoutlmv3, led, levit, lilt, llama, llama4, llama4_text, llava, llava_next, llava_next_video, llava_onevision, longformer, longt5, luke, lxmert, m2m_100, mamba, mamba2, marian, markuplm, mask2former, maskformer, maskformer-swin, mbart, mctct, mega, megatron-bert, mgp-str, mimi, mistral, mistral3, mixtral, mllama, mobilebert, mobilenet_v1, mobilenet_v2, mobilevit, mobilevitv2, modernbert, moonshine, moshi, mpnet, mpt, mra, mt5, musicgen, musicgen_melody, mvp, nat, nemotron, nezha, nllb-moe, nougat, nystromformer, olmo, olmo2, olmoe, omdet-turbo, oneformer, open-llama, openai-gpt, opt, owlv2, owlvit, paligemma, patchtsmixer, patchtst, pegasus, pegasus_x, perceiver, persimmon, phi, phi3, phi4_multimodal, phimoe, pix2struct, pixtral, plbart, poolformer, pop2piano, prompt_depth_anything, prophetnet, pvt, pvt_v2, qdqbert, qwen2, qwen2_5_vl, qwen2_audio, qwen2_audio_encoder, qwen2_moe, qwen2_vl, qwen3, qwen3_moe, rag, realm, recurrent_gemma, reformer, regnet, rembert, resnet, retribert, roberta, roberta-prelayernorm, roc_bert, roformer, rt_detr, rt_detr_resnet, rt_detr_v2, rwkv, sam, sam_vision_model, seamless_m4t, seamless_m4t_v2, segformer, seggpt, sew, sew-d, shieldgemma2, siglip, siglip2, siglip_vision_model, smolvlm, smolvlm_vision, speech-encoder-decoder, speech_to_text, speech_to_text_2, speecht5, splinter, squeezebert, stablelm, starcoder2, superglue, superpoint, swiftformer, swin, swin2sr, swinv2, switch_transformers, t5, table-transformer, tapas, textnet, time_series_transformer, timesformer, timm_backbone, timm_wrapper, trajectory_transformer, transfo-xl, trocr, tvlt, tvp, udop, umt5, unispeech, unispeech-sat, univnet, upernet, van, video_llava, videomae, vilt, vipllava, vision-encoder-decoder, vision-text-dual-encoder, visual_bert, vit, vit_hybrid, vit_mae, vit_msn, vitdet, vitmatte, vitpose, vitpose_backbone, vits, vivit, wav2vec2, wav2vec2-bert, wav2vec2-conformer, wavlm, whisper, xclip, xglm, xlm, xlm-prophetnet, xlm-roberta, xlm-roberta-xl, xlnet, xmod, yolos, yoso, zamba, zamba2, zoedepth
I found this very confusing…does anyone know what I am experiencing?
","Yea. Or maybe try reinstalling hf_transfer. If that’s the cause.
pip install -U hf_transfer hf_xet
+I have a COCO database in my disk (with a JSON in the annotations folder that contains image directions) and I would like to load it in HF dataset in orther to use CV models.
\nIs there a function that allows that?
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-06T12:13:56.138Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 283, 'reads': 9, 'readers_count': 8, 'score': 1381.8, 'yours': False, 'topic_id': 153752, 'topic_slug': 'load-a-coco-format-database-from-disk-for-detr', 'display_username': 'RAOUNAK LOUDAD', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/what-bounding-boxes-format-does-grounding-dino-use/161851/2', 'internal': True, 'reflection': True, 'title': 'What bounding boxes format does Grounding DINO use?', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93025, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/load-a-coco-format-database-from-disk-for-detr/153752/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 220222, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-07T01:56:39.463Z', 'cooked': 'Hmm… This?
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-07T01:56:39.463Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 9, 'readers_count': 8, 'score': 56.8, 'yours': False, 'topic_id': 153752, 'topic_slug': 'load-a-coco-format-database-from-disk-for-detr', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/datasets/issues/2526', 'internal': False, 'reflection': False, 'title': 'Add COCO datasets · Issue #2526 · huggingface/datasets · GitHub', 'clicks': 34}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/load-a-coco-format-database-from-disk-for-detr/153752/2', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 220344, 'name': 'Quentin Lhoest', 'username': 'lhoestq', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png', 'created_at': '2025-05-07T12:45:42.759Z', 'cooked': '\nThere is no COCO loader in the datasets library, but it would be a welcomed contribution in my opinion.
All the existing data modules are listed here
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-07T12:45:42.759Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 1, 'incoming_link_count': 11, 'reads': 6, 'readers_count': 5, 'score': 86.2, 'yours': False, 'topic_id': 153752, 'topic_slug': 'load-a-coco-format-database-from-disk-for-detr', 'display_username': 'Quentin Lhoest', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/datasets/tree/main/src/datasets/packaged_modules', 'internal': False, 'reflection': False, 'title': 'datasets/src/datasets/packaged_modules at main · huggingface/datasets · GitHub', 'clicks': 14}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 76, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/load-a-coco-format-database-from-disk-for-detr/153752/3', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221922, 'name': 'RAOUNAK LOUDAD', 'username': 'Godouche', 'avatar_template': '/user_avatar/discuss.huggingface.co/godouche/{size}/46990_2.png', 'created_at': '2025-05-14T12:48:46.156Z', 'cooked': 'I wrote this code for loading COCO datasets in hugging face datasets that works with DETR,
\nAdaptations:
\nimport json\nimport os\nimport subprocess\nfrom datasets import DatasetDict, Dataset, Features, Value, Sequence, ClassLabel, Image\n\n# Ensure the datasets module is installed\nsubprocess.check_call([""pip"", ""install"", ""datasets""])\n\nclass CocoDatasetLoader:\n def __init__(self, coco_folder):\n self.coco_folder = coco_folder\n\n def group_by_key_id(self, data, key_id, category_id_to_index):\n """"""\n Groups data by a specified key and maps category IDs to indices.\n \n Args:\n data (list): List of dictionaries containing the data.\n key_id (str): The key to group by.\n category_id_to_index (dict): Mapping from category IDs to indices.\n \n Returns:\n dict: Grouped data.\n """"""\n grouped_data = {}\n for item in data:\n key_value = item[key_id]\n if key_value not in grouped_data:\n grouped_data[key_value] = {k: [] for k in item.keys() if k != key_id}\n for k, v in item.items():\n if k != key_id:\n grouped_data[key_value][k].append(v)\n grouped_data[key_value][\'category\'] = [category_id_to_index[x] for x in grouped_data[key_value][\'category_id\']]\n return grouped_data\n \n def load_coco_hf_dataset(self, split):\n """"""\n Loads COCO dataset and processes it into a format suitable for Hugging Face datasets.\n \n Args:\n split (str): Dataset split (e.g., \'Train\', \'Test\', \'Validation\').\n \n Returns:\n Dataset: HuggingFace Dataset of the split of COCO dataset.\n """"""\n # Load the JSON file\n json_file_path = os.path.join(self.coco_folder, f\'annotations/instances_{split}.json\')\n try:\n with open(json_file_path, \'r\') as f:\n coco_data = json.load(f)\n except FileNotFoundError:\n print(f""File not found: {json_file_path}"")\n return []\n\n # Extract category names and create a mapping from category IDs to indices\n category_names = [cat[\'name\'] for cat in coco_data[\'categories\']]\n category_id_to_index = {cat[\'id\']: idx for idx, cat in enumerate(coco_data[\'categories\'])}\n\n # Group annotations by \'image_id\'\n grouped_annotations = self.group_by_key_id(coco_data[\'annotations\'], \'image_id\', category_id_to_index)\n\n # Create a dictionary of images\n grouped_images = {item[\'id\']: item for item in coco_data[\'images\']}\n\n # Initialize \'objects\' field in grouped_images\n annotations_keys = list(grouped_annotations.values())[0].keys()\n for k, v in grouped_images.items():\n grouped_images[k][\'objects\'] = {key: [] for key in annotations_keys}\n\n # Populate \'objects\' field with annotations\n for k, v in grouped_annotations.items():\n grouped_images[k][\'objects\'] = v\n\n # Add image paths and IDs\n for k, v in grouped_images.items():\n v[\'image\'] = os.path.join(self.coco_folder, \'images\', split, v[\'file_name\'])\n v[\'image_id\'] = v[\'id\']\n\n # Create a Hugging Face dataset from the custom data using from_list for efficiency\n hf_dataset = Dataset.from_list(list(grouped_images.values()))\n\n # Define the features for the main dataset\n features = Features({\n \'id\': Value(\'int64\'),\n \'image_id\': Value(\'int64\'),\n \'image\': Image(),\n \'file_name\': Value(\'string\'),\n \'license\': Value(\'string\'),\n \'flickr_url\': Value(\'string\'),\n \'coco_url\': Value(\'string\'),\n \'date_captured\': Value(\'string\'),\n \'width\': Value(\'int64\'),\n \'height\': Value(\'int64\'),\n \'objects\': Sequence({\n \'id\': Value(\'int64\'),\n \'area\': Value(\'float32\'),\n \'bbox\': Sequence(Value(\'float32\')),\n \'category\': ClassLabel(names=category_names),\n \'attributes\': {\'occluded\': Value(\'bool\')},\n \'category_id\': Value(\'int64\'),\n \'iscrowd\': Value(\'int64\'),\n \'segmentation\': {\n \'counts\': Sequence(Value(\'int64\')),\n \'size\': Sequence(Value(\'int64\'))\n }\n })\n })\n\n # Cast the features for the Hugging Face dataset\n hf_dataset = hf_dataset.cast(features)\n\n return hf_dataset\n\n# Initialize the CocoDatasetLoader class\ncoco_loader = CocoDatasetLoader(\'/path/to/coco/folder/\')\n\nhf_dataset_dict = DatasetDict()\nfor split in [\'Train\', \'Test\', \'Validation\']:\n # Load the COCO dataset for each split\n hf_dataset = coco_loader.load_coco_hf_dataset(split)\n \n # Print the dataset\n print(f""Dataset for {split} split:"")\n print(hf_dataset)\n \n # Create a DatasetDict with the split\n hf_dataset_dict[split.lower()] = hf_dataset\n\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-15T00:48:58.730Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 10.8, 'yours': False, 'topic_id': 153752, 'topic_slug': 'load-a-coco-format-database-from-disk-for-detr', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/load-a-coco-format-database-from-disk-for-detr/153752/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have a COCO database in my disk (with a JSON in the annotations folder that contains image directions) and I would like to load it in HF dataset in orther to use CV models.
+Is there a function that allows that?
","I wrote this code for loading COCO datasets in hugging face datasets that works with DETR,
+Adaptations:
+import json
+import os
+import subprocess
+from datasets import DatasetDict, Dataset, Features, Value, Sequence, ClassLabel, Image
+
+# Ensure the datasets module is installed
+subprocess.check_call([""pip"", ""install"", ""datasets""])
+
+class CocoDatasetLoader:
+ def __init__(self, coco_folder):
+ self.coco_folder = coco_folder
+
+ def group_by_key_id(self, data, key_id, category_id_to_index):
+ """"""
+ Groups data by a specified key and maps category IDs to indices.
+
+ Args:
+ data (list): List of dictionaries containing the data.
+ key_id (str): The key to group by.
+ category_id_to_index (dict): Mapping from category IDs to indices.
+
+ Returns:
+ dict: Grouped data.
+ """"""
+ grouped_data = {}
+ for item in data:
+ key_value = item[key_id]
+ if key_value not in grouped_data:
+ grouped_data[key_value] = {k: [] for k in item.keys() if k != key_id}
+ for k, v in item.items():
+ if k != key_id:
+ grouped_data[key_value][k].append(v)
+ grouped_data[key_value]['category'] = [category_id_to_index[x] for x in grouped_data[key_value]['category_id']]
+ return grouped_data
+
+ def load_coco_hf_dataset(self, split):
+ """"""
+ Loads COCO dataset and processes it into a format suitable for Hugging Face datasets.
+
+ Args:
+ split (str): Dataset split (e.g., 'Train', 'Test', 'Validation').
+
+ Returns:
+ Dataset: HuggingFace Dataset of the split of COCO dataset.
+ """"""
+ # Load the JSON file
+ json_file_path = os.path.join(self.coco_folder, f'annotations/instances_{split}.json')
+ try:
+ with open(json_file_path, 'r') as f:
+ coco_data = json.load(f)
+ except FileNotFoundError:
+ print(f""File not found: {json_file_path}"")
+ return []
+
+ # Extract category names and create a mapping from category IDs to indices
+ category_names = [cat['name'] for cat in coco_data['categories']]
+ category_id_to_index = {cat['id']: idx for idx, cat in enumerate(coco_data['categories'])}
+
+ # Group annotations by 'image_id'
+ grouped_annotations = self.group_by_key_id(coco_data['annotations'], 'image_id', category_id_to_index)
+
+ # Create a dictionary of images
+ grouped_images = {item['id']: item for item in coco_data['images']}
+
+ # Initialize 'objects' field in grouped_images
+ annotations_keys = list(grouped_annotations.values())[0].keys()
+ for k, v in grouped_images.items():
+ grouped_images[k]['objects'] = {key: [] for key in annotations_keys}
+
+ # Populate 'objects' field with annotations
+ for k, v in grouped_annotations.items():
+ grouped_images[k]['objects'] = v
+
+ # Add image paths and IDs
+ for k, v in grouped_images.items():
+ v['image'] = os.path.join(self.coco_folder, 'images', split, v['file_name'])
+ v['image_id'] = v['id']
+
+ # Create a Hugging Face dataset from the custom data using from_list for efficiency
+ hf_dataset = Dataset.from_list(list(grouped_images.values()))
+
+ # Define the features for the main dataset
+ features = Features({
+ 'id': Value('int64'),
+ 'image_id': Value('int64'),
+ 'image': Image(),
+ 'file_name': Value('string'),
+ 'license': Value('string'),
+ 'flickr_url': Value('string'),
+ 'coco_url': Value('string'),
+ 'date_captured': Value('string'),
+ 'width': Value('int64'),
+ 'height': Value('int64'),
+ 'objects': Sequence({
+ 'id': Value('int64'),
+ 'area': Value('float32'),
+ 'bbox': Sequence(Value('float32')),
+ 'category': ClassLabel(names=category_names),
+ 'attributes': {'occluded': Value('bool')},
+ 'category_id': Value('int64'),
+ 'iscrowd': Value('int64'),
+ 'segmentation': {
+ 'counts': Sequence(Value('int64')),
+ 'size': Sequence(Value('int64'))
+ }
+ })
+ })
+
+ # Cast the features for the Hugging Face dataset
+ hf_dataset = hf_dataset.cast(features)
+
+ return hf_dataset
+
+# Initialize the CocoDatasetLoader class
+coco_loader = CocoDatasetLoader('/path/to/coco/folder/')
+
+hf_dataset_dict = DatasetDict()
+for split in ['Train', 'Test', 'Validation']:
+ # Load the COCO dataset for each split
+ hf_dataset = coco_loader.load_coco_hf_dataset(split)
+
+ # Print the dataset
+ print(f""Dataset for {split} split:"")
+ print(hf_dataset)
+
+ # Create a DatasetDict with the split
+ hf_dataset_dict[split.lower()] = hf_dataset
+
+I have been averaging about 300-400 visits per week for a few months, but about a week ago new visits stopped registering and it shows no visits in the last week:
\n
However, my logs still show plenty of visitors using the space and I’ve had colleagues etc visit the site during the time frame without their visit being tracked and so it seems to be an issue with the tracking itself.
\nHas anyone else been noticing this issue? Relatively minor issue in the grand scheme of things but I have seen my place on the trending list completely fall off so it does seem to have some sort of effect that I’d like to fix if possible.
\nThanks!
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-12T04:43:13.613Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 43, 'reads': 7, 'readers_count': 6, 'score': 231.4, 'yours': False, 'topic_id': 154627, 'topic_slug': 'potential-issue-with-spaces-analytics-not-working', 'display_username': 'Nolan Zandi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91249, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/potential-issue-with-spaces-analytics-not-working/154627/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 221325, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-12T06:36:34.442Z', 'cooked': 'This seems like a bug… @pierric @Wauplin
\nIt seems that bug reports for Hub and Spaces can be submitted here.
Hi @nolanzandi thanks for reporting! We’re looking into it and I’ll update you soon.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-13T15:17:37.522Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 36.0, 'yours': False, 'topic_id': 154627, 'topic_slug': 'potential-issue-with-spaces-analytics-not-working', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/potential-issue-with-spaces-analytics-not-working/154627/3', 'reactions': [{'id': 'clap', 'type': 'emoji', 'count': 1}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221703, 'name': 'Nolan Zandi', 'username': 'nolanzandi', 'avatar_template': '/user_avatar/discuss.huggingface.co/nolanzandi/{size}/45859_2.png', 'created_at': '2025-05-13T16:11:19.467Z', 'cooked': 'Thank you so much @meganariley. I appreciate it!
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-13T16:11:19.467Z', 'reply_count': 1, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 154627, 'topic_slug': 'potential-issue-with-spaces-analytics-not-working', 'display_username': 'Nolan Zandi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 31941, 'username': 'meganariley', 'name': 'Megan Riley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91249, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/potential-issue-with-spaces-analytics-not-working/154627/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221864, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-05-14T09:38:49.608Z', 'cooked': 'Hi @nolanzandi thanks for waiting! This is now fixed. Let us know if you continue running into issues.
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-14T09:38:49.608Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 154627, 'topic_slug': 'potential-issue-with-spaces-analytics-not-working', 'display_username': 'Megan Riley', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 91249, 'username': 'nolanzandi', 'name': 'Nolan Zandi', 'avatar_template': '/user_avatar/discuss.huggingface.co/nolanzandi/{size}/45859_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 31941, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/potential-issue-with-spaces-analytics-not-working/154627/5', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 222085, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-14T21:39:45.766Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-05-14T21:39:45.766Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 1, 'readers_count': 0, 'score': 0.2, 'yours': False, 'topic_id': 154627, 'topic_slug': 'potential-issue-with-spaces-analytics-not-working', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/potential-issue-with-spaces-analytics-not-working/154627/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have been averaging about 300-400 visits per week for a few months, but about a week ago new visits stopped registering and it shows no visits in the last week:
+
However, my logs still show plenty of visitors using the space and I’ve had colleagues etc visit the site during the time frame without their visit being tracked and so it seems to be an issue with the tracking itself.
+Has anyone else been noticing this issue? Relatively minor issue in the grand scheme of things but I have seen my place on the trending list completely fall off so it does seem to have some sort of effect that I’d like to fix if possible.
+Thanks!
","Hi @nolanzandi thanks for waiting! This is now fixed. Let us know if you continue running into issues.
" +Is there any agent that can search google,https://discuss.huggingface.co/t/is-there-any-agent-that-can-search-google/141016,141016,25,2025-02-15 18:22:08.966000+00:00,"[{'id': 202756, 'name': 'elkahtib', 'username': 'Abdelkareem', 'avatar_template': '/user_avatar/discuss.huggingface.co/abdelkareem/{size}/30422_2.png', 'created_at': '2025-02-15T18:22:09.024Z', 'cooked': 'I want to build a smolagent that can search the results of google search ?
\nthere is the google API search but i don’t want to use it’s limit is very bad to me.
Setting up your own search engine for this task is more rewarding, and costs less.
\nThis is what I use for web search:
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-02-22T12:35:22.936Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 40, 'readers_count': 39, 'score': 93.0, 'yours': False, 'topic_id': 141016, 'topic_slug': 'is-there-any-agent-that-can-search-google', 'display_username': 'Michael Joiner', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/searxng/searxng', 'internal': False, 'reflection': False, 'title': 'GitHub - searxng/searxng: SearXNG is a free internet metasearch engine which aggregates results from various search services and databases. Users are neither tracked nor profiled.', 'clicks': 41}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 81771, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/is-there-any-agent-that-can-search-google/141016/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 205862, 'name': 'gael', 'username': 'gael1130', 'avatar_template': '/user_avatar/discuss.huggingface.co/gael1130/{size}/42164_2.png', 'created_at': '2025-02-28T10:40:19.048Z', 'cooked': 'Yes, you can use the GoogleSearchTool, which is one of the default tools of smolagents.
\nimport os\nfrom smolagents import GoogleSearchTool, HfApiModel\nos.environ[""SERPAPI_API_KEY""] = userdata.get(\'SERPAPI_API_KEY\')\n\nmodel = HfApiModel(model_id=""Qwen/Qwen2.5-Coder-32B-Instruct"", provider=""together"")\n\nagent = CodeAgent(\n model=model,\n tools=[GoogleSearchTool()]\n)\nThe link to get your Serp API key.
\nAnd if you want to go beyond, you can use the DuckDuckGoSearchTool. It also has limits but maybe a combination of both can help?
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-05-13T12:09:37.100Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 11, 'readers_count': 10, 'score': 7.2, 'yours': False, 'topic_id': 141016, 'topic_slug': 'is-there-any-agent-that-can-search-google', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/is-there-any-agent-that-can-search-google/141016/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I want to build a smolagent that can search the results of google search ?
+there is the google API search but i don’t want to use it’s limit is very bad to me.
Yes, you can use the GoogleSearchTool, which is one of the default tools of smolagents.
+import os
+from smolagents import GoogleSearchTool, HfApiModel
+os.environ[""SERPAPI_API_KEY""] = userdata.get('SERPAPI_API_KEY')
+
+model = HfApiModel(model_id=""Qwen/Qwen2.5-Coder-32B-Instruct"", provider=""together"")
+
+agent = CodeAgent(
+ model=model,
+ tools=[GoogleSearchTool()]
+)
+The link to get your Serp API key.
+And if you want to go beyond, you can use the DuckDuckGoSearchTool. It also has limits but maybe a combination of both can help?
I am trying to create a simple langchain app on text-generation using API to communicate with models on HuggingFace servers.
\nI created a “.env” file and stored by KEY in the variable: “HUGGINGFACEHUB_API_TOKEN”
\nI also checked it, API token is valid.
Post that, I tried running this code snippet:
\n from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint\n from dotenv import load_dotenv\n\n load_dotenv()\n\n llm = HuggingFaceEndpoint(\n repo_id=""TinyLlama/TinyLlama-1.1B-Chat-v1.0"",\n task=""text-generation""\n )\n\n model = ChatHuggingFace(llm=llm)\n result = model.invoke(""What is the capital of India"")\n print(result.content)\nThis is giving an error. I tried multiple things around it, but nothing worked.
\nHere is the error log:
\nTraceback (most recent call last):
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\2.ChatModels\\2_chatmodel_hf_api.py”, line 13, in
\nresult = model.invoke(“What is the capital of India”)
\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\langchain_core\\language_models\\chat_models.py”, line 370, in invoke
\nself.generate_prompt(
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\langchain_core\\language_models\\chat_models.py”, line 947, in generate_prompt
\nreturn self.generate(prompt_messages, stop=stop, callbacks=callbacks, **kwargs)
\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\langchain_core\\language_models\\chat_models.py”, line 766, in generate
\nself._generate_with_cache(
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\langchain_core\\language_models\\chat_models.py”, line 1012, in _generate_with_cache
\nresult = self._generate(
\n^^^^^^^^^^^^^^^
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\langchain_huggingface\\chat_models\\huggingface.py”, line 574, in generate
\nanswer = self.llm.client.chat_completion(messages=message_dicts, **params)
\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
\nFile “C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\huggingface_hub\\inference_client.py”, line 886, in chat_completion
\nprovider_helper = get_provider_helper(
\n^^^^^^^^^^^^^^^^^^^^
\nFile ""C:\\Users\\SS\\Desktop\\Camp_langchain_models\\venv\\Lib\\site-packages\\huggingface_hub\\inference_providers_init.py"", line 165, in get_provider_helper
\nprovider = next(iter(provider_mapping))
\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^
\nStopIteration
I am new to it. Any guidance around this is much appreciated. Thank you.
', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-05-11T09:15:16.322Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 420, 'reads': 37, 'readers_count': 36, 'score': 2107.4, 'yours': False, 'topic_id': 154529, 'topic_slug': 'facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key', 'display_username': 'S', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/stopiteration-error/155463/2', 'internal': True, 'reflection': True, 'title': 'Stopiteration error', 'clicks': 7}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93574, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key/154529/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 221179, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-11T10:04:01.158Z', 'cooked': 'I think LangChain has not yet caught up with the changes in Hugging Face’s specifications.
\n\n\n\nMeanwhile, one possible solution would be to downgrade your
\nhuggingface-hubversion to 0.27.1 or below.
pip install huggingface_hub<=0.27.1\nI am also facing similar issue
\nplease let me know if you found any solution
pip install langchain-huggingface langchain
from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint\nllm = HuggingFaceEndpoint(\n repo_id=""deepseek-ai/DeepSeek-R1"",\n provider=""together""\n)\nmodel = ChatHuggingFace(llm=llm)\nresult = model.invoke(""What is the capital of India"")\nThis works for me with the following setup:
\n$ pip freeze | grep huggingface\nhuggingface-hub==0.31.1\nlangchain-huggingface==0.2.0\n$ pip freeze | grep langchain\nlangchain==0.3.25\nlangchain-core==0.3.59\nlangchain-huggingface==0.2.0\nlangchain-text-splitters==0.3.8\nPlease note the following regarding TinyLlama/TinyLlama-1.1B-Chat-v1.0:
\n\n', 'post_number': 5, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-05-11T16:12:40.609Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 32, 'readers_count': 31, 'score': 61.4, 'yours': False, 'topic_id': 154529, 'topic_slug': 'facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key', 'display_username': 'Mahmut C', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0', 'internal': False, 'reflection': False, 'title': 'TinyLlama/TinyLlama-1.1B-Chat-v1.0 · Hugging Face', 'clicks': 20}], 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 61570, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key/154529/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221221, 'name': 'S', 'username': 'Shaleensr', 'avatar_template': '/user_avatar/discuss.huggingface.co/shaleensr/{size}/47299_2.png', 'created_at': '2025-05-11T16:25:46.336Z', 'cooked': 'This model isn’t deployed by any Inference Provider.
\n
Thank you @mahmutc. This code snippet worked for me.
', 'post_number': 6, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-05-11T16:25:46.336Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 29, 'readers_count': 28, 'score': 25.8, 'yours': False, 'topic_id': 154529, 'topic_slug': 'facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key', 'display_username': 'S', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 61570, 'username': 'mahmutc', 'name': 'Mahmut C', 'avatar_template': '/user_avatar/discuss.huggingface.co/mahmutc/{size}/52583_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93574, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key/154529/6', 'reactions': [{'id': 'confetti_ball', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221222, 'name': 'S', 'username': 'Shaleensr', 'avatar_template': '/user_avatar/discuss.huggingface.co/shaleensr/{size}/47299_2.png', 'created_at': '2025-05-11T16:28:01.145Z', 'cooked': 'The below snippet by mahmutc worked for me:
\n', 'post_number': 7, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-05-11T16:28:01.145Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 1, 'incoming_link_count': 5, 'reads': 29, 'readers_count': 28, 'score': 45.8, 'yours': False, 'topic_id': 154529, 'topic_slug': 'facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key', 'display_username': 'S', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 93503, 'username': 'niteshburnwal', 'name': 'NITESH KUMAR', 'avatar_template': '/user_avatar/discuss.huggingface.co/niteshburnwal/{size}/47260_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93574, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key/154529/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 221312, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-12T04:28:01.352Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 8, 'updated_at': '2025-05-12T04:28:01.352Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 20, 'readers_count': 19, 'score': 29.0, 'yours': False, 'topic_id': 154529, 'topic_slug': 'facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/facing-issue-using-a-model-hosted-on-huggingface-server-and-talking-to-it-using-api-key/154529/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am trying to create a simple langchain app on text-generation using API to communicate with models on HuggingFace servers.
+I created a “.env” file and stored by KEY in the variable: “HUGGINGFACEHUB_API_TOKEN”
+I also checked it, API token is valid.
Post that, I tried running this code snippet:
+ from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
+ from dotenv import load_dotenv
+
+ load_dotenv()
+
+ llm = HuggingFaceEndpoint(
+ repo_id=""TinyLlama/TinyLlama-1.1B-Chat-v1.0"",
+ task=""text-generation""
+ )
+
+ model = ChatHuggingFace(llm=llm)
+ result = model.invoke(""What is the capital of India"")
+ print(result.content)
+This is giving an error. I tried multiple things around it, but nothing worked.
+Here is the error log:
+Traceback (most recent call last):
+File “C:\Users\SS\Desktop\Camp_langchain_models\2.ChatModels\2_chatmodel_hf_api.py”, line 13, in
+result = model.invoke(“What is the capital of India”)
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+File “C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\langchain_core\language_models\chat_models.py”, line 370, in invoke
+self.generate_prompt(
+File “C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\langchain_core\language_models\chat_models.py”, line 947, in generate_prompt
+return self.generate(prompt_messages, stop=stop, callbacks=callbacks, **kwargs)
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+File “C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\langchain_core\language_models\chat_models.py”, line 766, in generate
+self._generate_with_cache(
+File “C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\langchain_core\language_models\chat_models.py”, line 1012, in _generate_with_cache
+result = self._generate(
+^^^^^^^^^^^^^^^
+File “C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\langchain_huggingface\chat_models\huggingface.py”, line 574, in generate
+answer = self.llm.client.chat_completion(messages=message_dicts, **params)
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+File “C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\huggingface_hub\inference_client.py”, line 886, in chat_completion
+provider_helper = get_provider_helper(
+^^^^^^^^^^^^^^^^^^^^
+File ""C:\Users\SS\Desktop\Camp_langchain_models\venv\Lib\site-packages\huggingface_hub\inference_providers_init.py"", line 165, in get_provider_helper
+provider = next(iter(provider_mapping))
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+StopIteration
I am new to it. Any guidance around this is much appreciated. Thank you.
","pip install langchain-huggingface langchain
from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
+llm = HuggingFaceEndpoint(
+ repo_id=""deepseek-ai/DeepSeek-R1"",
+ provider=""together""
+)
+model = ChatHuggingFace(llm=llm)
+result = model.invoke(""What is the capital of India"")
+This works for me with the following setup:
+$ pip freeze | grep huggingface
+huggingface-hub==0.31.1
+langchain-huggingface==0.2.0
+$ pip freeze | grep langchain
+langchain==0.3.25
+langchain-core==0.3.59
+langchain-huggingface==0.2.0
+langchain-text-splitters==0.3.8
+I am a student currently working on training the LLAMA-4-Scout-17B-16E-Instruct model using LoRA, running on an H100 GPU with 80GB VRAM (on Lambda Labs). However, I have encountered an out of memory error during the training process. I understand that this might fall slightly outside the scope of the course, but despite extensive research and reviewing various community discussions, I have not been able to resolve the issue.
\nHere is a brief outline of my setup:
\nHardware: H100 (80GB VRAM)
\nModel: LLAMA-4-Scout-17B-16E-Instruct (download on unsloth hugging face)
\nTraining Method: LoRA
\nError: CUDA out of memory
\nCode snippet:
\nimport torch
\nfrom transformers import AutoTokenizer,TrainingArguments,Trainer,DataCollatorForLanguageModeling,AutoModelForCausalLM
\nfrom peft import LoraConfig, get_peft_model, TaskType
\nfrom datasets import load_dataset
\nfrom accelerate import dispatch_model
\nfrom accelerate import Accelerator
\nfrom accelerate.utils import get_balanced_memory, infer_auto_device_map
\nimport os
\nos.environ[“PYTORCH_CUDA_ALLOC_CONF”] = “expandable_segments:True”
model_path = “/home/ubuntu/llama4”
\ndataset_path = “llama_nc_instruction_train.jsonl”
\noutput_dir = “./merged_llama4_nccode”
print(“ loading tokenizer…”)
\ntokenizer = AutoTokenizer.from_pretrained(model_path)
print(“ loading model…（使用 safetensors）”)
\nmodel = AutoModelForCausalLM.from_pretrained(
\nmodel_path,
\ntorch_dtype=torch.bfloat16,
\nlow_cpu_mem_usage=True,
\ntrust_remote_code=True
\n)
print(“ applying LoRA setting…”)
\nlora_config = LoraConfig(
\nr=8,
\nlora_alpha=32, #有人用8
\ntarget_modules=[“q_proj”, “v_proj”],
\nlora_dropout=0.05,
\nbias=“none”,
\ntask_type=TaskType.CAUSAL_LM,
\n)
model = get_peft_model(model, lora_config)
\nprint(“ loading data…”)
\ndataset = load_dataset(“json”, data_files=dataset_path, split=“train”)
def tokenize(example):
\ntokenized_inputs = tokenizer(
\nexample[“text”],
\ntruncation=True,
\npadding=“max_length”,
\nmax_length=4196
\n)
\nreturn tokenized_inputs
tokenized_dataset = dataset.map(tokenize, batched=True, remove_columns=[“text”])
\nprint(“ establish Trainer…”)
\ntraining_args = TrainingArguments(
\noutput_dir=“./lora_tmp”,
\nnum_train_epochs=3,
\nper_device_train_batch_size=1, #有人用64
\ngradient_accumulation_steps=512,
\nlearning_rate=2e-4,
\nlogging_steps=10,
\nsave_strategy=“no”,
\n)
trainer = Trainer(
\nmodel=model,
\nargs=training_args,
\ntrain_dataset=tokenized_dataset,
\ntokenizer=tokenizer,
\ndata_collator=DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False),
\n)
print(“ training…”)
\ntrainer.train()
print(“ merge LoRA weight…”)
\nmodel = model.merge_and_unload()
print(“ save model to:”, output_dir)
\nmodel.save_pretrained(output_dir)
\ntokenizer.save_pretrained(output_dir)
print(“ finish！”)
and this is the error:
\n 載入 tokenizer…
\n 載入模型…（使用 safetensors）
\nLoading checkpoint shards: 100%|███████████████████████████████████████████████████████| 50/50 [00:00<00:00, 457.56it/s]
\n 套用 LoRA 設定…
\n 載入資料中…
\n 建立 Trainer…
\n/home/ubuntu/CNC代碼定義訓練黨TEST.py:68: FutureWarning: tokenizer is deprecated and will be removed in version 5.0.0 for Trainer.init. Use processing_class instead.
\ntrainer = Trainer(
\nTraceback (most recent call last):
\nFile “/home/ubuntu/CNC代碼定義訓練黨TEST.py”, line 68, in
\ntrainer = Trainer(
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/transformers/utils/deprecation.py”, line 172, in wrapped_func
\nreturn func(*args, **kwargs)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/transformers/trainer.py”, line 614, in init
\nself._move_model_to_device(model, args.device)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/transformers/trainer.py”, line 901, in _move_model_to_device
\nmodel = model.to(device)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 1355, in to
\nreturn self._apply(convert)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 915, in _apply
\nmodule._apply(fn)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 915, in _apply
\nmodule._apply(fn)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 915, in _apply
\nmodule._apply(fn)
\n[Previous line repeated 4 more times]
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 942, in _apply
\nparam_applied = fn(param)
\nFile “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 1341, in convert
\nreturn t.to(
\ntorch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.25 GiB. GPU 0 has a total capacity of 79.19 GiB of which 359.06 MiB is free. Including non-PyTorch memory, this process has 78.83 GiB memory in use. Of the allocated memory 78.38 GiB is allocated by PyTorch, and 8.21 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (CUDA semantics — PyTorch 2.7 documentation)
Would anyone kindly offer any suggestions or best practices to address this issue? Are there specific parameters I should consider adjusting (e.g., batch size, gradient checkpointing, LoRA rank, etc.) to make it fit within the memory constraints?
\nOr is this simply a case of hardware limitation, and even 80GB VRAM is not enough for this model.And i have tried the QLORA method,encountering the same question.
It looks like you’re running into a CUDA out of memory issue while fine-tuning LLAMA-4-Scout-17B-16E-Instruct using LoRA on an H100 GPU with 80GB VRAM. Even though 80GB is a lot, large models like this can still exceed memory limits, especially with high batch sizes and gradient accumulation steps.
\nper_device_train_batch_size=1, your gradient_accumulation_steps=512 might be causing excessive memory usage.r=8) and target modules (q_proj, v_proj) might be consuming more memory than expected.max_length=4196 is quite large, leading to high memory consumption per sample.PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True, fragmentation might still be an issue.Try lowering gradient_accumulation_steps to 128 or 64 instead of 512:
training_args = TrainingArguments(\n output_dir=""./lora_tmp"",\n num_train_epochs=3,\n per_device_train_batch_size=1,\n gradient_accumulation_steps=64, # Reduce from 512\n learning_rate=2e-4,\n logging_steps=10,\n save_strategy=""no"",\n)\nThis will reduce memory usage significantly.
\nTry reducing max_length from 4196 to 2048:
tokenized_inputs = tokenizer(\n example[""text""],\n truncation=True,\n padding=""max_length"",\n max_length=2048 # Reduce from 4196\n)\nThis will cut memory usage per sample in half.
\nThis helps reduce memory usage by recomputing activations instead of storing them:
\nmodel.gradient_checkpointing_enable()\ntorch.compile() for OptimizationIf you’re using PyTorch 2.0+, try compiling the model for better memory efficiency:
\nmodel = torch.compile(model)\nIf memory is still an issue, offload parts of the model to CPU using accelerate:
from accelerate import infer_auto_device_map, dispatch_model\n\ndevice_map = infer_auto_device_map(model, max_memory={""cuda"": ""75GB"", ""cpu"": ""20GB""})\nmodel = dispatch_model(model, device_map=device_map)\nThis ensures that only essential parts stay on the GPU.
\nTry these adjustments one by one and monitor memory usage. If the issue persists, consider switching to QLoRA with 4-bit quantization, which significantly reduces VRAM usage.
\nLet me know if you need help implementing these fixes!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-09T15:08:51.365Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 4, 'readers_count': 3, 'score': 15.6, 'yours': False, 'topic_id': 153432, 'topic_slug': 'inquiry-regarding-out-of-memory-issue-during-lora-fine-tuning', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/inquiry-regarding-out-of-memory-issue-during-lora-fine-tuning/153432/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am a student currently working on training the LLAMA-4-Scout-17B-16E-Instruct model using LoRA, running on an H100 GPU with 80GB VRAM (on Lambda Labs). However, I have encountered an out of memory error during the training process. I understand that this might fall slightly outside the scope of the course, but despite extensive research and reviewing various community discussions, I have not been able to resolve the issue.
+Here is a brief outline of my setup:
+Hardware: H100 (80GB VRAM)
+Model: LLAMA-4-Scout-17B-16E-Instruct (download on unsloth hugging face)
+Training Method: LoRA
+Error: CUDA out of memory
+Code snippet:
+import torch
+from transformers import AutoTokenizer,TrainingArguments,Trainer,DataCollatorForLanguageModeling,AutoModelForCausalLM
+from peft import LoraConfig, get_peft_model, TaskType
+from datasets import load_dataset
+from accelerate import dispatch_model
+from accelerate import Accelerator
+from accelerate.utils import get_balanced_memory, infer_auto_device_map
+import os
+os.environ[“PYTORCH_CUDA_ALLOC_CONF”] = “expandable_segments:True”
model_path = “/home/ubuntu/llama4”
+dataset_path = “llama_nc_instruction_train.jsonl”
+output_dir = “./merged_llama4_nccode”
print(“ loading tokenizer…”)
+tokenizer = AutoTokenizer.from_pretrained(model_path)
print(“ loading model…（使用 safetensors）”)
+model = AutoModelForCausalLM.from_pretrained(
+model_path,
+torch_dtype=torch.bfloat16,
+low_cpu_mem_usage=True,
+trust_remote_code=True
+)
print(“ applying LoRA setting…”)
+lora_config = LoraConfig(
+r=8,
+lora_alpha=32, #有人用8
+target_modules=[“q_proj”, “v_proj”],
+lora_dropout=0.05,
+bias=“none”,
+task_type=TaskType.CAUSAL_LM,
+)
model = get_peft_model(model, lora_config)
+print(“ loading data…”)
+dataset = load_dataset(“json”, data_files=dataset_path, split=“train”)
def tokenize(example):
+tokenized_inputs = tokenizer(
+example[“text”],
+truncation=True,
+padding=“max_length”,
+max_length=4196
+)
+return tokenized_inputs
tokenized_dataset = dataset.map(tokenize, batched=True, remove_columns=[“text”])
+print(“ establish Trainer…”)
+training_args = TrainingArguments(
+output_dir=“./lora_tmp”,
+num_train_epochs=3,
+per_device_train_batch_size=1, #有人用64
+gradient_accumulation_steps=512,
+learning_rate=2e-4,
+logging_steps=10,
+save_strategy=“no”,
+)
trainer = Trainer(
+model=model,
+args=training_args,
+train_dataset=tokenized_dataset,
+tokenizer=tokenizer,
+data_collator=DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False),
+)
print(“ training…”)
+trainer.train()
print(“ merge LoRA weight…”)
+model = model.merge_and_unload()
print(“ save model to:”, output_dir)
+model.save_pretrained(output_dir)
+tokenizer.save_pretrained(output_dir)
print(“ finish！”)
and this is the error:
+ 載入 tokenizer…
+ 載入模型…（使用 safetensors）
+Loading checkpoint shards: 100%|███████████████████████████████████████████████████████| 50/50 [00:00<00:00, 457.56it/s]
+ 套用 LoRA 設定…
+ 載入資料中…
+ 建立 Trainer…
+/home/ubuntu/CNC代碼定義訓練黨TEST.py:68: FutureWarning: tokenizer is deprecated and will be removed in version 5.0.0 for Trainer.init. Use processing_class instead.
+trainer = Trainer(
+Traceback (most recent call last):
+File “/home/ubuntu/CNC代碼定義訓練黨TEST.py”, line 68, in
+trainer = Trainer(
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/transformers/utils/deprecation.py”, line 172, in wrapped_func
+return func(*args, **kwargs)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/transformers/trainer.py”, line 614, in init
+self._move_model_to_device(model, args.device)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/transformers/trainer.py”, line 901, in _move_model_to_device
+model = model.to(device)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 1355, in to
+return self._apply(convert)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 915, in _apply
+module._apply(fn)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 915, in _apply
+module._apply(fn)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 915, in _apply
+module._apply(fn)
+[Previous line repeated 4 more times]
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 942, in _apply
+param_applied = fn(param)
+File “/home/ubuntu/llama_env/lib/python3.10/site-packages/torch/nn/modules/module.py”, line 1341, in convert
+return t.to(
+torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.25 GiB. GPU 0 has a total capacity of 79.19 GiB of which 359.06 MiB is free. Including non-PyTorch memory, this process has 78.83 GiB memory in use. Of the allocated memory 78.38 GiB is allocated by PyTorch, and 8.21 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (CUDA semantics — PyTorch 2.7 documentation)
Would anyone kindly offer any suggestions or best practices to address this issue? Are there specific parameters I should consider adjusting (e.g., batch size, gradient checkpointing, LoRA rank, etc.) to make it fit within the memory constraints?
+Or is this simply a case of hardware limitation, and even 80GB VRAM is not enough for this model.And i have tried the QLORA method,encountering the same question.
It looks like you’re running into a CUDA out of memory issue while fine-tuning LLAMA-4-Scout-17B-16E-Instruct using LoRA on an H100 GPU with 80GB VRAM. Even though 80GB is a lot, large models like this can still exceed memory limits, especially with high batch sizes and gradient accumulation steps.
+per_device_train_batch_size=1, your gradient_accumulation_steps=512 might be causing excessive memory usage.r=8) and target modules (q_proj, v_proj) might be consuming more memory than expected.max_length=4196 is quite large, leading to high memory consumption per sample.PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True, fragmentation might still be an issue.Try lowering gradient_accumulation_steps to 128 or 64 instead of 512:
training_args = TrainingArguments(
+ output_dir=""./lora_tmp"",
+ num_train_epochs=3,
+ per_device_train_batch_size=1,
+ gradient_accumulation_steps=64, # Reduce from 512
+ learning_rate=2e-4,
+ logging_steps=10,
+ save_strategy=""no"",
+)
+This will reduce memory usage significantly.
+Try reducing max_length from 4196 to 2048:
tokenized_inputs = tokenizer(
+ example[""text""],
+ truncation=True,
+ padding=""max_length"",
+ max_length=2048 # Reduce from 4196
+)
+This will cut memory usage per sample in half.
+This helps reduce memory usage by recomputing activations instead of storing them:
+model.gradient_checkpointing_enable()
+torch.compile() for OptimizationIf you’re using PyTorch 2.0+, try compiling the model for better memory efficiency:
+model = torch.compile(model)
+If memory is still an issue, offload parts of the model to CPU using accelerate:
from accelerate import infer_auto_device_map, dispatch_model
+
+device_map = infer_auto_device_map(model, max_memory={""cuda"": ""75GB"", ""cpu"": ""20GB""})
+model = dispatch_model(model, device_map=device_map)
+This ensures that only essential parts stay on the GPU.
+Try these adjustments one by one and monitor memory usage. If the issue persists, consider switching to QLoRA with 4-bit quantization, which significantly reduces VRAM usage.
+Let me know if you need help implementing these fixes!
Hello everyone I am very new and im experimenting with the Huggingface Autotrain UI but im having a little trouble getting the training started. I am trying to train a meta-llama/Llama-3.1-8b-Instruct Model with an example dataset that i found
\nalpaca1k.csv
\nwhich i uploaded as a local file.
\nI have not made any changes to any other parameters. When i then click start training i get an error.
ERROR | 2025-05-08 07:39:20 | autotrain.trainers.common:wrapper:215 - train has failed due to an exception: Traceback (most recent call last):
\nFile “/app/env/lib/python3.10/site-packages/autotrain/trainers/common.py”, line 212, in wrapper
\nreturn func(*args, **kwargs)
\nFile “/app/env/lib/python3.10/site-packages/autotrain/trainers/clm/main.py”, line 28, in train
\ntrain_sft(config)
\nFile “/app/env/lib/python3.10/site-packages/autotrain/trainers/clm/train_clm_sft.py”, line 27, in train
\nmodel = utils.get_model(config, tokenizer)
\nFile “/app/env/lib/python3.10/site-packages/autotrain/trainers/clm/utils.py”, line 943, in get_model
\nmodel = AutoModelForCausalLM.from_pretrained(
\nFile “/app/env/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py”, line 564, in from_pretrained
\nreturn model_class.from_pretrained(
\nFile “/app/env/lib/python3.10/site-packages/transformers/modeling_utils.py”, line 3620, in from_pretrained
\nhf_quantizer.validate_environment(
\nFile “/app/env/lib/python3.10/site-packages/transformers/quantizers/quantizer_bnb_4bit.py”, line 83, in validate_environment
\nvalidate_bnb_backend_availability(raise_exception=True)
\nFile “/app/env/lib/python3.10/site-packages/transformers/integrations/bitsandbytes.py”, line 559, in validate_bnb_backend_availability
\nreturn _validate_bnb_cuda_backend_availability(raise_exception)
\nFile “/app/env/lib/python3.10/site-packages/transformers/integrations/bitsandbytes.py”, line 537, in _validate_bnb_cuda_backend_availability
\nraise RuntimeError(log_msg)
\nRuntimeError: CUDA is required but not available for bitsandbytes. Please consider installing the multi-platform enabled version of bitsandbytes, which is currently a work in progress. Please check currently supported platforms and installation instructions at Installation Guide
ERROR | 2025-05-08 07:39:20 | autotrain.trainers.common:wrapper:216 - CUDA is required but not available for bitsandbytes. Please consider installing the multi-platform enabled version of bitsandbytes, which is currently a work in progress. Please check currently supported platforms and installation instructions at Installation Guide
\nINFO | 2025-05-08 07:39:20 | autotrain.trainers.common:pause_space:156 - Pausing space…
I not sure how i can fix this. Any help is appreciated
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-08T07:41:32.922Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 231, 'reads': 11, 'readers_count': 10, 'score': 1147.2, 'yours': False, 'topic_id': 154069, 'topic_slug': 'error-in-autotrain-training', 'display_username': 'Lukas', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/bitsandbytes/main/en/installation#multi-backend', 'internal': False, 'reflection': False, 'title': 'Installation Guide', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93248, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-in-autotrain-training/154069/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 220527, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-08T08:06:56.954Z', 'cooked': 'In some cases, the problem can be resolved by installing bitsandbytes as indicated in the error message. However, in other cases, reinstalling PyTorch and the CUDA Toolkit may be necessary.
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-08T08:06:56.954Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 8, 'readers_count': 7, 'score': 16.6, 'yours': False, 'topic_id': 154069, 'topic_slug': 'error-in-autotrain-training', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/bitsandbytes-foundation/bitsandbytes/issues/1093', 'internal': False, 'reflection': False, 'title': 'RuntimeError: Failed to import transformers.integrations.bitsandbytes because of the following error (look up to see its traceback): · Issue #1093 · bitsandbytes-foundation/bitsandbytes · GitHub', 'clicks': 8}, {'url': 'https://github.com/bitsandbytes-foundation/bitsandbytes/issues/1384', 'internal': False, 'reflection': False, 'title': 'An error occurred: CUDA is required but not available for bitsandbytes. · Issue #1384 · bitsandbytes-foundation/bitsandbytes · GitHub', 'clicks': 6}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-in-autotrain-training/154069/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 220532, 'name': 'Lukas', 'username': 'LuuWee', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/l/4af34b/{size}.png', 'created_at': '2025-05-08T08:17:02.201Z', 'cooked': 'I found a solution by myself. Im using the free plan to there is only cpu to use and no gpu. I had to change some of the parameters. This is what i did for anyone who is wondering
\nDistributed Backend from ddp to deepspeed
\nMixed precision from fp16 to none
\nPEFT/LoRA from true to false
Im not exactly sure what did the trick but its training now
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-08T08:17:02.201Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 7, 'readers_count': 6, 'score': 41.4, 'yours': False, 'topic_id': 154069, 'topic_slug': 'error-in-autotrain-training', 'display_username': 'Lukas', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93248, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/error-in-autotrain-training/154069/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 220669, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-08T20:17:56.235Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-05-08T20:17:56.235Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 7, 'reads': 5, 'readers_count': 4, 'score': 36.0, 'yours': False, 'topic_id': 154069, 'topic_slug': 'error-in-autotrain-training', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/error-in-autotrain-training/154069/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello everyone I am very new and im experimenting with the Huggingface Autotrain UI but im having a little trouble getting the training started. I am trying to train a meta-llama/Llama-3.1-8b-Instruct Model with an example dataset that i found
+alpaca1k.csv
+which i uploaded as a local file.
+I have not made any changes to any other parameters. When i then click start training i get an error.
ERROR | 2025-05-08 07:39:20 | autotrain.trainers.common:wrapper:215 - train has failed due to an exception: Traceback (most recent call last):
+File “/app/env/lib/python3.10/site-packages/autotrain/trainers/common.py”, line 212, in wrapper
+return func(*args, **kwargs)
+File “/app/env/lib/python3.10/site-packages/autotrain/trainers/clm/main.py”, line 28, in train
+train_sft(config)
+File “/app/env/lib/python3.10/site-packages/autotrain/trainers/clm/train_clm_sft.py”, line 27, in train
+model = utils.get_model(config, tokenizer)
+File “/app/env/lib/python3.10/site-packages/autotrain/trainers/clm/utils.py”, line 943, in get_model
+model = AutoModelForCausalLM.from_pretrained(
+File “/app/env/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py”, line 564, in from_pretrained
+return model_class.from_pretrained(
+File “/app/env/lib/python3.10/site-packages/transformers/modeling_utils.py”, line 3620, in from_pretrained
+hf_quantizer.validate_environment(
+File “/app/env/lib/python3.10/site-packages/transformers/quantizers/quantizer_bnb_4bit.py”, line 83, in validate_environment
+validate_bnb_backend_availability(raise_exception=True)
+File “/app/env/lib/python3.10/site-packages/transformers/integrations/bitsandbytes.py”, line 559, in validate_bnb_backend_availability
+return _validate_bnb_cuda_backend_availability(raise_exception)
+File “/app/env/lib/python3.10/site-packages/transformers/integrations/bitsandbytes.py”, line 537, in _validate_bnb_cuda_backend_availability
+raise RuntimeError(log_msg)
+RuntimeError: CUDA is required but not available for bitsandbytes. Please consider installing the multi-platform enabled version of bitsandbytes, which is currently a work in progress. Please check currently supported platforms and installation instructions at Installation Guide
ERROR | 2025-05-08 07:39:20 | autotrain.trainers.common:wrapper:216 - CUDA is required but not available for bitsandbytes. Please consider installing the multi-platform enabled version of bitsandbytes, which is currently a work in progress. Please check currently supported platforms and installation instructions at Installation Guide
+INFO | 2025-05-08 07:39:20 | autotrain.trainers.common:pause_space:156 - Pausing space…
I not sure how i can fix this. Any help is appreciated
","I found a solution by myself. Im using the free plan to there is only cpu to use and no gpu. I had to change some of the parameters. This is what i did for anyone who is wondering
+Distributed Backend from ddp to deepspeed
+Mixed precision from fp16 to none
+PEFT/LoRA from true to false
Im not exactly sure what did the trick but its training now
" +Join the Hugging Face Discord!,https://discuss.huggingface.co/t/join-the-hugging-face-discord/11263,11263,12,2021-11-01 15:54:32.137000+00:00,"[{'id': 24338, 'name': 'Nate Raw', 'username': 'nateraw', 'avatar_template': '/user_avatar/discuss.huggingface.co/nateraw/{size}/2556_2.png', 'created_at': '2021-11-01T15:54:32.206Z', 'cooked': 'We’re excited to announce our official community discord server! We will have community events, sprints, reading clubs and more! Here’s the link to join: http://hf.co/join/discord
So, any questions should still be directed here.
From looking at the HTML, it seems that that is an empty link. I know it’s November 1st, but aren’t jokes for April 1st?
For future visitors who like to click instead of type, here you go.
', 'post_number': 2, 'post_type': 1, 'posts_count': 41, 'updated_at': '2022-04-08T07:23:29.676Z', 'reply_count': 2, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 11, 'reads': 369, 'readers_count': 368, 'score': 183.8, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Bram Vanroy', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'http://hf.co/join/discord', 'internal': False, 'reflection': False, 'title': 'Hugging Face', 'clicks': 478}], 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 23, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 3}, {'id': 'hugs', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 24344, 'name': 'Nate Raw', 'username': 'nateraw', 'avatar_template': '/user_avatar/discuss.huggingface.co/nateraw/{size}/2556_2.png', 'created_at': '2021-11-01T17:51:02.459Z', 'cooked': 'whoops, nice catch! I used markdown syntax to add the link, but it didn’t go through not sure what’s up with that. Anyways, I fixed the link in the original post too. Thanks, Bram
hello everyone,
\nI present my self, I’m Mohamed BEN ALI research engineer.
\nI want to join hugging face community via Discord.
\nThanks
The link has expired. Mind sharing a new one? thanks!
', 'post_number': 5, 'post_type': 1, 'posts_count': 41, 'updated_at': '2022-04-08T02:29:43.263Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 23, 'reads': 181, 'readers_count': 180, 'score': 156.2, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Teoh Sin Yee', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 23, 'username': 'BramVanroy', 'name': 'Bram Vanroy', 'avatar_template': '/user_avatar/discuss.huggingface.co/bramvanroy/{size}/47360_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 7117, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 34053, 'name': 'Nate Raw', 'username': 'nateraw', 'avatar_template': '/user_avatar/discuss.huggingface.co/nateraw/{size}/2556_2.png', 'created_at': '2022-04-08T02:54:17.808Z', 'cooked': 'The link in the original post should still be working
The discord invite here and in HF website is invalid. At least it is the message that appear for me.
', 'post_number': 7, 'post_type': 1, 'posts_count': 41, 'updated_at': '2022-09-29T12:40:12.921Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 17, 'reads': 119, 'readers_count': 118, 'score': 108.8, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Fred Guth', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 4558, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/7', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 48823, 'name': 'Nate Raw', 'username': 'nateraw', 'avatar_template': '/user_avatar/discuss.huggingface.co/nateraw/{size}/2556_2.png', 'created_at': '2022-11-07T18:39:30.512Z', 'cooked': 'I know this response is very late, but this link still works as far as I can tell may have been out temporarily when you replied @fredguth
I finally did my post for all three. Cool HF space on Discord @nateraw is there any way or future where I can integrate a space and allow AI input/output onto a Discord chat channel or server? I’ve been infatuated with Mid Journey interface on Discord lately as a neat jam session way to multiplayer access to AI in real time. Super excited to see what you are cooking up. --Aaron
', 'post_number': 9, 'post_type': 1, 'posts_count': 41, 'updated_at': '2022-12-03T12:40:50.288Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 26, 'reads': 107, 'readers_count': 106, 'score': 151.4, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Aaron C Wacker', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6987, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/9', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 84953, 'name': 'Carlos', 'username': 'nbalive', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/n/e68b1a/{size}.png', 'created_at': '2023-08-19T02:05:40.166Z', 'cooked': 'The invite is invalid for me
The invite link (Hugging Face) doesn’t work for me - I just see ‘Unable to accept invite’.
', 'post_number': 11, 'post_type': 1, 'posts_count': 41, 'updated_at': '2023-09-22T19:57:43.823Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 63, 'readers_count': 62, 'score': 47.6, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Pat Patterson', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/join/discord', 'internal': False, 'reflection': False, 'title': 'Hugging Face', 'clicks': 12}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 29597, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/11', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 91128, 'name': 'Radamés Ajna', 'username': 'radames', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png', 'created_at': '2023-09-22T22:11:00.940Z', 'cooked': 'hi @metadaddy, I jus tested the link Hugging Face and seems to be working. @lunarflu could you please check?
', 'post_number': 12, 'post_type': 1, 'posts_count': 41, 'updated_at': '2023-09-22T22:11:00.940Z', 'reply_count': 1, 'reply_to_post_number': 11, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 63, 'readers_count': 62, 'score': 37.6, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Radamés Ajna', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discord.com/invite/JfAtkvEtRb', 'internal': False, 'reflection': False, 'title': 'Hugging Face', 'clicks': 20}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 29597, 'username': 'metadaddy', 'name': 'Pat Patterson', 'avatar_template': '/user_avatar/discuss.huggingface.co/metadaddy/{size}/52440_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 6306, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/12', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 91130, 'name': 'Pat Patterson', 'username': 'metadaddy', 'avatar_template': '/user_avatar/discuss.huggingface.co/metadaddy/{size}/52440_2.png', 'created_at': '2023-09-22T22:49:34.239Z', 'cooked': 'Hi @radames - I figured it out - Discord needs to be running for the invitation process to work correctly. If it’s not, then you get the ‘unable to accept invite’ message, rather than any advice to start Discord.
\nThanks!
', 'post_number': 13, 'post_type': 1, 'posts_count': 41, 'updated_at': '2023-09-22T22:49:34.239Z', 'reply_count': 1, 'reply_to_post_number': 12, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 63, 'readers_count': 62, 'score': 87.6, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'Pat Patterson', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 6306, 'username': 'radames', 'name': 'Radamés Ajna', 'avatar_template': '/user_avatar/discuss.huggingface.co/radames/{size}/28246_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 29597, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/13', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 91234, 'name': 'Adam Molnar', 'username': 'lunarflu', 'avatar_template': '/user_avatar/discuss.huggingface.co/lunarflu/{size}/29357_2.png', 'created_at': '2023-09-23T17:29:24.291Z', 'cooked': 'Happy to hear that. Enjoy, and share your thoughts with the world!
I can’t join, why?
', 'post_number': 16, 'post_type': 1, 'posts_count': 41, 'updated_at': '2024-09-19T10:45:48.832Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 57, 'readers_count': 56, 'score': 41.4, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'mamat mamation', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 64844, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/16', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 156210, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2024-09-19T10:51:38.322Z', 'cooked': '@nateraw The HF Discord key posted on the HF Forum appears to have expired.
', 'post_number': 17, 'post_type': 1, 'posts_count': 41, 'updated_at': '2024-09-19T10:51:38.322Z', 'reply_count': 1, 'reply_to_post_number': 16, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 68, 'readers_count': 67, 'score': 63.6, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 64844, 'username': 'mmty', 'name': 'mamat mamation', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/m/dfb087/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/17', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 159113, 'name': 'Adam Molnar', 'username': 'lunarflu', 'avatar_template': '/user_avatar/discuss.huggingface.co/lunarflu/{size}/29357_2.png', 'created_at': '2024-09-30T10:26:31.510Z', 'cooked': 'Hey @John6666 @mmty ! Feel free to try this link, or alternatively, you can try searching hugging face within Discord. Let me know if it works!
\n
Thanks for the update. But I don’t have a Discord account so I’ll leave it to someone else!
I was able to unearth an ancient, unused Discord account, so I joined!
', 'post_number': 21, 'post_type': 1, 'posts_count': 41, 'updated_at': '2024-10-29T04:41:13.879Z', 'reply_count': 1, 'reply_to_post_number': 19, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 47, 'readers_count': 46, 'score': 59.4, 'yours': False, 'topic_id': 11263, 'topic_slug': 'join-the-hugging-face-discord', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'reply_to_user': {'id': 15783, 'username': 'lunarflu', 'name': 'Adam Molnar', 'avatar_template': '/user_avatar/discuss.huggingface.co/lunarflu/{size}/29357_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/join-the-hugging-face-discord/11263/21', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 168305, 'name': 'Edward Surridge', 'username': 'EdSurridge', 'avatar_template': '/user_avatar/discuss.huggingface.co/edsurridge/{size}/34137_2.png', 'created_at': '2024-11-07T11:40:21.424Z', 'cooked': 'I am interested to join what you found . Thanks if you can share it
\nEd
We’re excited to announce our official community discord server! We will have community events, sprints, reading clubs and more! Here’s the link to join: http://hf.co/join/discord
So, any questions should still be directed here.
I am interested to join what you found . Thanks if you can share it
+Ed
Hi,
\nIThe following code snippet for pulling a pretrained custom tokenizer from the Hugging Face Hub
\nimport os\nfrom transformers import AutoTokenizer\n\n# load the tokenizer\ntokenizer = AutoTokenizer.from_pretrained(""smostafanejad/gen-mlm-cismi-bert-wordpiece"",\n token=os.environ[\'HF_TOKEN\'],\n cache_dir=""./cache""\n )\nsuddenly started raising the following runtime error since yesterday (05/05/2025).
\nCell In[4], line 5\n 2 from transformers import AutoTokenizer\n 4 # load the tokenizer\n----> 5 tokenizer = AutoTokenizer.from_pretrained(""smostafanejad/gen-mlm-cismi-bert-wordpiece"",\n 6 token=os.environ[\'HF_TOKEN\'],\n 7 cache_dir=""./cache""\n 8 )\n\nFile ~/Packages/miniconda3/envs/bertchemai/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py:992, in AutoTokenizer.from_pretrained(cls, pretrained_model_name_or_path, *inputs, **kwargs)\n 989 tokenizer_class_py, tokenizer_class_fast = TOKENIZER_MAPPING[type(config)]\n 991 if tokenizer_class_fast and (use_fast or tokenizer_class_py is None):\n--> 992 return tokenizer_class_fast.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)\n 993 else:\n 994 if tokenizer_class_py is not None:\n\nFile ~/Packages/miniconda3/envs/bertchemai/lib/python3.10/site-packages/transformers/tokenization_utils_base.py:2046, in PreTrainedTokenizerBase.from_pretrained(cls, pretrained_model_name_or_path, cache_dir, force_download, local_files_only, token, revision, trust_remote_code, *init_inputs, **kwargs)\n 2043 # If one passes a GGUF file path to `gguf_file` there is no need for this check as the tokenizer will be\n 2044 # loaded directly from the GGUF file.\n 2045 if all(full_file_name is None for full_file_name in resolved_vocab_files.values()) and not gguf_file:\n-> 2046 raise EnvironmentError(\n 2047 f""Can\'t load tokenizer for \'{pretrained_model_name_or_path}\'. If you were trying to load it from ""\n 2048 ""\'https://huggingface.co/models\', make sure you don\'t have a local directory with the same name. ""\n 2049 f""Otherwise, make sure \'{pretrained_model_name_or_path}\' is the correct path to a directory ""\n 2050 f""containing all relevant files for a {cls.__name__} tokenizer.""\n 2051 )\n 2053 for file_id, file_path in vocab_files.items():\n 2054 if file_id not in resolved_vocab_files:\n\nOSError: Can\'t load tokenizer for \'smostafanejad/gen-mlm-cismi-bert-wordpiece\'. If you were trying to load it from \'https://huggingface.co/models\', make sure you don\'t have a local directory with the same name. Otherwise, make sure \'smostafanejad/gen-mlm-cismi-bert-wordpiece\' is the correct path to a directory containing all relevant files for a BertTokenizerFast tokenizer.\nI have followed the suggestions in the error message (directory is clean and the address on the Hub is available) but they do not help.
\nI appreciate any assistance on this matter as the same function call used to work until yesterday.
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-06T19:41:08.528Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 164, 'reads': 12, 'readers_count': 11, 'score': 822.4, 'yours': False, 'topic_id': 153809, 'topic_slug': 'autotokenizer-from-pretrained-suddenly-raises-an-error', 'display_username': 'Sina Mostafanejad', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 70171, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/autotokenizer-from-pretrained-suddenly-raises-an-error/153809/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 220194, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-06T23:18:34.825Z', 'cooked': 'Hmm, it seems to be working. Maybe it’s a problem specific to ipython or Jupyter, or maybe it was a bug that occurred when you upgraded Transoformers. Or maybe it’s a network problem?
\nimport os\nfrom transformers import AutoTokenizer\n\n# load the tokenizer\ntokenizer = AutoTokenizer.from_pretrained(""smostafanejad/gen-mlm-cismi-bert-wordpiece"",\n #token=os.environ[\'HF_TOKEN\'],\n #cache_dir=""./cache""\n )\nprint(tokenizer)\n""""""\nPreTrainedTokenizerFast(name_or_path=\'smostafanejad/gen-mlm-cismi-bert-wordpiece\', vocab_size=30522, model_max_length=512, is_fast=True, padding_side=\'right\', truncation_side=\'right\', special_tokens={\'unk_token\': \'[UNK]\', \'sep_token\': \'[SEP]\', \'pad_token\': \'[PAD]\', \'cls_token\': \'[CLS]\', \'mask_token\': \'[MASK]\'}, clean_up_tokenization_spaces=False, added_tokens_decoder={\n 0: AddedToken(""[PAD]"", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 1: AddedToken(""[UNK]"", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 2: AddedToken(""[CLS]"", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 3: AddedToken(""[SEP]"", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 4: AddedToken(""[MASK]"", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n}\n)\n""""""\nYou are right and the problem does not seem to be related to Jupyter or ipython either.
\n\nI have now two machines with conda environments that suddenly started generating errors without doing anything to them. My personal laptop with a fresh conda environment seem to be fine (as you can see in the screenshot). So, I exported the problematic and OK conda environments and uploaded them to the repo to see if I can find out what’s causing the issue:
\nThanks for the time you’ve taken and tested the function call, @John6666.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-07T03:02:04.783Z', 'reply_count': 1, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 6, 'readers_count': 5, 'score': 36.2, 'yours': False, 'topic_id': 153809, 'topic_slug': 'autotokenizer-from-pretrained-suddenly-raises-an-error', 'display_username': 'Sina Mostafanejad', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/smostafanejad/gen-mlm-cismi-bert-wordpiece/blob/main/good_env.yml', 'internal': False, 'reflection': False, 'title': 'good_env.yml · smostafanejad/gen-mlm-cismi-bert-wordpiece at main', 'clicks': 2}, {'url': 'https://huggingface.co/smostafanejad/gen-mlm-cismi-bert-wordpiece/blob/main/bad_env.yml', 'internal': False, 'reflection': False, 'title': 'bad_env.yml · smostafanejad/gen-mlm-cismi-bert-wordpiece at main', 'clicks': 1}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 70171, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/autotokenizer-from-pretrained-suddenly-raises-an-error/153809/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 220377, 'name': 'Sina Mostafanejad', 'username': 'smostafanejad', 'avatar_template': '/user_avatar/discuss.huggingface.co/smostafanejad/{size}/34306_2.png', 'created_at': '2025-05-07T14:39:35.439Z', 'cooked': 'OK since this was an EnvironmentError I checked everything and I think I have found the culprit.
\nIn my bashrc, I had export HF_HUB_ENABLE_HF_TRANSFER=1 set which means the problem might have something to do with an inconsistency with the hf-transfer package. Reading Hugging Face’s Environment Variable documentation gave the clue about the possibility of such incidents and undefined behavior
HF_HUB_ENABLE_HF_TRANSFER\n\nSet to True to download files from the Hub using hf_transfer. It’s a Rust-based package that enables faster download (up to x2 speed-up). Be aware that this is still experimental so it might cause issues in your workflow. In particular, it does not support features such as progress bars, resume download, proxies or error handling.\n\nNote: hf_transfer has to be installed separately from Pypi.\nI have forced a reinstall and upgrade through pip and apparently that resolved the issues with both supercomputer and data center machines which had problems calling the AutoTokenizer.from_pretrained().
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-08T02:40:20.217Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 1.0, 'yours': False, 'topic_id': 153809, 'topic_slug': 'autotokenizer-from-pretrained-suddenly-raises-an-error', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/autotokenizer-from-pretrained-suddenly-raises-an-error/153809/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+IThe following code snippet for pulling a pretrained custom tokenizer from the Hugging Face Hub
+import os
+from transformers import AutoTokenizer
+
+# load the tokenizer
+tokenizer = AutoTokenizer.from_pretrained(""smostafanejad/gen-mlm-cismi-bert-wordpiece"",
+ token=os.environ['HF_TOKEN'],
+ cache_dir=""./cache""
+ )
+suddenly started raising the following runtime error since yesterday (05/05/2025).
+Cell In[4], line 5
+ 2 from transformers import AutoTokenizer
+ 4 # load the tokenizer
+----> 5 tokenizer = AutoTokenizer.from_pretrained(""smostafanejad/gen-mlm-cismi-bert-wordpiece"",
+ 6 token=os.environ['HF_TOKEN'],
+ 7 cache_dir=""./cache""
+ 8 )
+
+File ~/Packages/miniconda3/envs/bertchemai/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py:992, in AutoTokenizer.from_pretrained(cls, pretrained_model_name_or_path, *inputs, **kwargs)
+ 989 tokenizer_class_py, tokenizer_class_fast = TOKENIZER_MAPPING[type(config)]
+ 991 if tokenizer_class_fast and (use_fast or tokenizer_class_py is None):
+--> 992 return tokenizer_class_fast.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)
+ 993 else:
+ 994 if tokenizer_class_py is not None:
+
+File ~/Packages/miniconda3/envs/bertchemai/lib/python3.10/site-packages/transformers/tokenization_utils_base.py:2046, in PreTrainedTokenizerBase.from_pretrained(cls, pretrained_model_name_or_path, cache_dir, force_download, local_files_only, token, revision, trust_remote_code, *init_inputs, **kwargs)
+ 2043 # If one passes a GGUF file path to `gguf_file` there is no need for this check as the tokenizer will be
+ 2044 # loaded directly from the GGUF file.
+ 2045 if all(full_file_name is None for full_file_name in resolved_vocab_files.values()) and not gguf_file:
+-> 2046 raise EnvironmentError(
+ 2047 f""Can't load tokenizer for '{pretrained_model_name_or_path}'. If you were trying to load it from ""
+ 2048 ""'https://huggingface.co/models', make sure you don't have a local directory with the same name. ""
+ 2049 f""Otherwise, make sure '{pretrained_model_name_or_path}' is the correct path to a directory ""
+ 2050 f""containing all relevant files for a {cls.__name__} tokenizer.""
+ 2051 )
+ 2053 for file_id, file_path in vocab_files.items():
+ 2054 if file_id not in resolved_vocab_files:
+
+OSError: Can't load tokenizer for 'smostafanejad/gen-mlm-cismi-bert-wordpiece'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure 'smostafanejad/gen-mlm-cismi-bert-wordpiece' is the correct path to a directory containing all relevant files for a BertTokenizerFast tokenizer.
+I have followed the suggestions in the error message (directory is clean and the address on the Hub is available) but they do not help.
+I appreciate any assistance on this matter as the same function call used to work until yesterday.
","OK since this was an EnvironmentError I checked everything and I think I have found the culprit.
+In my bashrc, I had export HF_HUB_ENABLE_HF_TRANSFER=1 set which means the problem might have something to do with an inconsistency with the hf-transfer package. Reading Hugging Face’s Environment Variable documentation gave the clue about the possibility of such incidents and undefined behavior
HF_HUB_ENABLE_HF_TRANSFER
+
+Set to True to download files from the Hub using hf_transfer. It’s a Rust-based package that enables faster download (up to x2 speed-up). Be aware that this is still experimental so it might cause issues in your workflow. In particular, it does not support features such as progress bars, resume download, proxies or error handling.
+
+Note: hf_transfer has to be installed separately from Pypi.
+I have forced a reinstall and upgrade through pip and apparently that resolved the issues with both supercomputer and data center machines which had problems calling the AutoTokenizer.from_pretrained().
Hi there,
\nI am trying to figure out where documentation for models exists. For example, I am looking at the pipeline documentation which says that task is some id. But it is not a user defined id because passing “foo” as the task to the model gemma-3-27b-it gives me an error that lists all the tasks. Is there a function to call that lists the tasks ahead of time without having to get an error message? It is not clear from the documentation that the tasks are implemented by each model not the pipeline api - and it would be nice to know exactly what a model does for implementation purposes rather than some generic description of tasks in the tutorial. I would rather have some way of figuring out what a particular model does so I can implement it. Are there any tools that help me figure this out? Maybe it’s possible to parse it from the config files or the model file?
Also, how can I get information on message formatting for each task? Is there a way to figure this out or is it dependent on the information provided on the model card? So if the tasks and message formats are not listed on the model card, is there a way to determine these? Especially because I am also not seeing any source code implementing a model class that lists tasks and message parsers. Maybe there is a way to parse these from the config or model files as well?
\nThanks
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-26T02:21:47.120Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 22, 'reads': 12, 'readers_count': 11, 'score': 122.4, 'yours': False, 'topic_id': 152365, 'topic_slug': 'can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does', 'display_username': 'Sven Voigt', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/transformers/v4.51.3/en/main_classes/pipelines#transformers.Pipeline', 'internal': False, 'reflection': False, 'title': 'Pipelines', 'clicks': 1}, {'url': 'https://huggingface.co/google/gemma-3-27b-it', 'internal': False, 'reflection': False, 'title': 'google/gemma-3-27b-it · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91985, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does/152365/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 218318, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-26T08:44:58.165Z', 'cooked': 'It seems that tasks are being retrieved from classes registered in AutoModel, so you should be able to identify the problem by checking whether the class corresponding to the task is defined in the code.
\nI’m not sure if there is a simple method (a dedicated function) for this…
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-26T08:44:58.165Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 11, 'readers_count': 10, 'score': 2.2, 'yours': False, 'topic_id': 152365, 'topic_slug': 'can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/blob/main/src/transformers/pipelines/__init__.py#L877', 'internal': False, 'reflection': False, 'title': 'transformers/src/transformers/pipelines/__init__.py at main · huggingface/transformers · GitHub', 'clicks': 0}, {'url': 'https://github.com/huggingface/transformers/blob/main/src/transformers/models/auto/modeling_auto.py', 'internal': False, 'reflection': False, 'title': 'transformers/src/transformers/models/auto/modeling_auto.py at main · huggingface/transformers · GitHub', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does/152365/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218524, 'name': 'Sven Voigt', 'username': 'svenpvoigt', 'avatar_template': '/user_avatar/discuss.huggingface.co/svenpvoigt/{size}/46353_2.png', 'created_at': '2025-04-27T18:32:02.143Z', 'cooked': '@John6666 Thanks that’s a good place to start looking!
\nAlso, to add an example to the original post, the jinaai-embeddings model implements all custom tasks but lists them on the model card (e.g., retrieval.query, text-matching). However, it is unclear what the input format should be for each task just from the model card. It looks like lists of strings, but would need to see the model implementation to be sure there aren’t other options.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-27T18:32:24.674Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 152365, 'topic_slug': 'can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does', 'display_username': 'Sven Voigt', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/jinaai/jina-embeddings-v3', 'internal': False, 'reflection': False, 'title': 'jinaai/jina-embeddings-v3 · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91985, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does/152365/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 220179, 'name': 'Sven Voigt', 'username': 'svenpvoigt', 'avatar_template': '/user_avatar/discuss.huggingface.co/svenpvoigt/{size}/46353_2.png', 'created_at': '2025-05-06T22:42:54.575Z', 'cooked': 'I think I have an answer:
\nthe message format is always a list of strings for the tokenizer, unless the tokenizer includes a template. In that case the template can be printed out with tokenizer.chat_template and usually includes system and user roles as well as some keywords like add_generation_prompt.
However, it doesn’t seem to be overall standardized and there is a lot of custom code for models.
\nSo final answer: most everything has to be explained in the model card and you have to kind of figure out how to make it work from a couple examples.
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-06T22:42:54.575Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 152365, 'topic_slug': 'can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does', 'display_username': 'Sven Voigt', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91985, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does/152365/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 220314, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-07T10:43:41.493Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-07T10:43:41.493Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 152365, 'topic_slug': 'can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/can-i-get-clarification-on-what-exactly-transformers-does-vs-what-the-model-does/152365/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi there,
+I am trying to figure out where documentation for models exists. For example, I am looking at the pipeline documentation which says that task is some id. But it is not a user defined id because passing “foo” as the task to the model gemma-3-27b-it gives me an error that lists all the tasks. Is there a function to call that lists the tasks ahead of time without having to get an error message? It is not clear from the documentation that the tasks are implemented by each model not the pipeline api - and it would be nice to know exactly what a model does for implementation purposes rather than some generic description of tasks in the tutorial. I would rather have some way of figuring out what a particular model does so I can implement it. Are there any tools that help me figure this out? Maybe it’s possible to parse it from the config files or the model file?
Also, how can I get information on message formatting for each task? Is there a way to figure this out or is it dependent on the information provided on the model card? So if the tasks and message formats are not listed on the model card, is there a way to determine these? Especially because I am also not seeing any source code implementing a model class that lists tasks and message parsers. Maybe there is a way to parse these from the config or model files as well?
+Thanks
","I think I have an answer:
+the message format is always a list of strings for the tokenizer, unless the tokenizer includes a template. In that case the template can be printed out with tokenizer.chat_template and usually includes system and user roles as well as some keywords like add_generation_prompt.
However, it doesn’t seem to be overall standardized and there is a lot of custom code for models.
+So final answer: most everything has to be explained in the model card and you have to kind of figure out how to make it work from a couple examples.
" +403 Error: “Private repository storage limit reached” — quota shows space remaining,https://discuss.huggingface.co/t/403-error-private-repository-storage-limit-reached-quota-shows-space-remaining/153121,153121,23,2025-05-01 12:19:13.054000+00:00,"[{'id': 219303, 'name': 'Théo Boyer', 'username': 'Theob', 'avatar_template': '/user_avatar/discuss.huggingface.co/theob/{size}/30775_2.png', 'created_at': '2025-05-01T12:19:13.110Z', 'cooked': 'Hi,
\nI’m getting the following error when trying to push to my private dataset repo using huggingface_hub:
403 Forbidden: Private repository storage limit reached, please upgrade your plan...\nHowever, when I check my organization quota on the Hugging Face UI, it shows we’re only using ~66 GB out of the 100 GB available.
\nAny advice on how to find the root cause of this discrepancy ?
\nThanks!
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-01T12:19:13.110Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 83, 'reads': 18, 'readers_count': 17, 'score': 423.4, 'yours': False, 'topic_id': 153121, 'topic_slug': '403-error-private-repository-storage-limit-reached-quota-shows-space-remaining', 'display_username': 'Théo Boyer', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/under-500-mb-in-storage-but-indicates-1-gb/166347/2', 'internal': True, 'reflection': True, 'title': 'Under 500 MB in storage, but indicates 1 GB', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 2}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 30390, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/403-error-private-repository-storage-limit-reached-quota-shows-space-remaining/153121/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 2, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 219312, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-01T12:45:00.165Z', 'cooked': 'There is a phenomenon where past git commit entries accumulate and waste space, but even in that case, the size itself should be displayed in the settings screen. This phenomenon is probably an error or a bad specification. @meganariley @pierric
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-01T12:45:00.165Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 15, 'readers_count': 14, 'score': 37.8, 'yours': False, 'topic_id': 153121, 'topic_slug': '403-error-private-repository-storage-limit-reached-quota-shows-space-remaining', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/hub/storage-limits', 'internal': False, 'reflection': False, 'title': 'Storage limits', 'clicks': 4}, {'url': 'https://discuss.huggingface.co/t/spaces-force-push-getting-repository-storage-limit-reached/130269', 'internal': True, 'reflection': False, 'title': 'Spaces force push getting ""Repository storage limit reached""', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/403-error-private-repository-storage-limit-reached-quota-shows-space-remaining/153121/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 219768, 'name': 'Andrew J tokar', 'username': 'Zelgodiz', 'avatar_template': '/user_avatar/discuss.huggingface.co/zelgodiz/{size}/45662_2.png', 'created_at': '2025-05-05T04:30:01.968Z', 'cooked': 'It looks like you’re encountering a quota discrepancy issue on Hugging Face, where your storage limit error doesn’t match the actual usage shown in the UI. This has been reported by other users as well43dcd9a7-70db-4a1f-b0ae-981daa162054.
\nfrom huggingface_hub import HfApi\napi = HfApi()\nlfs_files = list(api.list_lfs_files(repo_id=""your_repo"", repo_type=""dataset""))\ntotal_size = sum(file.size for file in lfs_files)\nprint(f""Total LFS storage used: {total_size / (1024**3)} GB"")\ngit lfs prune\nLet me know if you need help troubleshooting further!
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-05-06T21:38:42.706Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 6, 'readers_count': 5, 'score': 1.0, 'yours': False, 'topic_id': 153121, 'topic_slug': '403-error-private-repository-storage-limit-reached-quota-shows-space-remaining', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/403-error-private-repository-storage-limit-reached-quota-shows-space-remaining/153121/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi,
+I’m getting the following error when trying to push to my private dataset repo using huggingface_hub:
403 Forbidden: Private repository storage limit reached, please upgrade your plan...
+However, when I check my organization quota on the Hugging Face UI, it shows we’re only using ~66 GB out of the 100 GB available.
+Any advice on how to find the root cause of this discrepancy ?
+Thanks!
","" +Prepare dataset from YOLO format to COCO for DETR,https://discuss.huggingface.co/t/prepare-dataset-from-yolo-format-to-coco-for-detr/34894,34894,9,2023-03-28 10:19:48.796000+00:00,"[{'id': 62739, 'name': 'Alberto Ruiz', 'username': 'Alberto1404', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/a/90ced4/{size}.png', 'created_at': '2023-03-28T10:19:48.868Z', 'cooked': 'Hi. I would like to compare two nets using the same dataset, regardless being Transformer-based (DETR) vs Non-Transformer based (YOLOv5).
\nI have already trained a model using Yolov5, such that my dataset is already split into train-val-test, in YOLO format. See Formatting table to visualize an example. My dataset folder looks like this:
.\n├── train\n └── images\n │ ├── ima1.png\n │ ├── ima2.png\n │ ├── ...\n └── labels\n │ ├── ima1.txt\n │ ├── ima2.txt\n │ ├── ...\n├── val\n └── images\n │ ├── ima3.png\n │ ├── ima4.png\n │ ├── ...\n └── labels\n │ ├── ima3.txt\n │ ├── ima4.txt\n │ ├── ...\n├── test\n └── images\n │ ├── ima5.png\n │ ├── ima6.png\n │ ├── ...\n └── labels\n │ ├── ima5.txt\n │ ├── ima6.txt\n │ ├── ...\nNow I want to convert it to COCO format. From Hugging Face documentation, DETR demands COCO format in labels, using JSON files. However, you are using a dataset loaded from Hugging Face datasets library. Moreover, I would like to know if I should create 3 JSON files, for each split, or 1 JSON file containing all. In the latter case, could you provide some documentation on how should the JSON file be defined?
\nIf there is any tutorial on how to prepare the data to feed DETR, based on my specs, it would be nice to post it here.
\nThank you for all!
I did the following parser to convert it.
\nimport os\nimport json\nfrom PIL import Image\nfrom tqdm import tqdm\n\n\ndef yolo_to_coco(image_dir, label_dir, output_dir):\n\t# Define categories\n\tcategories = [{\'id\': 0, \'name\': \'person\'}]\n\n\t# Initialize data dict\n\tdata = {\'train\': [], \'validation\': [], \'test\': []}\n\n\t# Loop over splits\n\tfor split in [\'train\', \'validation\', \'test\']:\n\t\tsplit_data = {\'info\': {}, \'licenses\': [], \'images\': [], \'annotations\': [], \'categories\': categories}\n\n\t\t# Get image and label files for current split\n\t\timage_files = sorted(os.listdir(image_dir))\n\t\tlabel_files = sorted(os.listdir(label_dir))\n\n\t\t# Loop over images in current split\n\t\tcumulative_id = 0\n\t\twith tqdm(total=len(image_files), desc=f\'Processing {split} images\') as pbar:\n\t\t\tfor i, filename in enumerate(image_files):\n\t\t\t\timage_path = os.path.join(image_dir, filename)\n\t\t\t\tim = Image.open(image_path)\n\t\t\t\tim_id = i + 1\n\n\t\t\t\tsplit_data[\'images\'].append({\n\t\t\t\t\t\'id\': im_id,\n\t\t\t\t\t\'file_name\': filename,\n\t\t\t\t\t\'width\': im.size[0],\n\t\t\t\t\t\'height\': im.size[1]\n\t\t\t\t})\n\n\t\t\t\t# Get labels for current image\n\t\t\t\tlabel_path = os.path.join(label_dir, os.path.splitext(filename)[0] + \'.txt\')\n\t\t\t\twith open(label_path, \'r\') as f:\n\t\t\t\t\tyolo_data = f.readlines()\n\n\t\t\t\tfor line in yolo_data:\n\t\t\t\t\tclass_id, x_center, y_center, width, height = line.split()\n\t\t\t\t\tclass_id = int(class_id)\n\t\t\t\t\tbbox_x = (float(x_center) - float(width) / 2) * im.size[0]\n\t\t\t\t\tbbox_y = (float(y_center) - float(height) / 2) * im.size[1]\n\t\t\t\t\tbbox_width = float(width) * im.size[0]\n\t\t\t\t\tbbox_height = float(height) * im.size[1]\n\n\t\t\t\t\tsplit_data[\'annotations\'].append({\n\t\t\t\t\t\t\'id\': cumulative_id,\n\t\t\t\t\t\t\'image_id\': im_id,\n\t\t\t\t\t\t\'category_id\': class_id,\n\t\t\t\t\t\t\'bbox\': [bbox_x, bbox_y, bbox_width, bbox_height],\n\t\t\t\t\t\t\'area\': bbox_width * bbox_height,\n\t\t\t\t\t\t\'iscrowd\': 0\n\t\t\t\t\t})\n\n\t\t\t\t\tcumulative_id += 1\n\n\t\t\t\tpbar.update(1)\n\n\t\tdata[split] = split_data\n\n\t# Save data to JSON files\n\tfor split in [\'train\', \'validation\', \'test\']:\n\t\tfilename = os.path.join(output_dir, f\'{split}.json\')\n\t\twith open(filename, \'w\') as f:\n\t\t\tjson.dump({\'data\': data[split]}, f)\n\n\treturn data\n\nimage_dir = \'/home/alberto/Dataset/train/images\'\nlabel_dir = \'/home/alberto/Dataset/train/labels\'\noutput_dir = \'./\'\ncoco_data = yolo_to_coco(image_dir, label_dir, output_dir)\n\nHowever, when I want to load my dataset using:
\nfrom datasets import load_dataset\ndata_files = {\n\t""train"": \'/home/alberto/Dataset/train/images/train_labels.json\',\n\t""validation"": \'/home/alberto/Dataset/val/images/val_labels.json\',\n\t""test"": \'/home/alberto/Dataset/val/images/test_labels.json\'\n}\ndataset = load_dataset(""json"", data_files=data_files)\nTyping dataset[\'train\'] outputs that number of rows is 1, which is not correct. It should be 7000, the number of images in the train set. Does anybody know where the error is commited?
\nExample with subset of train set:
\n
In order to read it using load_dataset, it is a must to follow the same structure as defined
\nhere
@Alberto1404 Have you find out the final script to convert from yolo format to coco for DETR? Have you resolved this issue"" typing dataset[\'train\'] outputs that number of rows is 1, which is not correct. It should be 7000, the number of images in the train set. Does anybody know where the error is commited?""
could you please provide the solution to transform YOLO to COCO for DETR?
', 'post_number': 5, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-05-06T12:03:48.957Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 34894, 'topic_slug': 'prepare-dataset-from-yolo-format-to-coco-for-detr', 'display_username': 'RAOUNAK LOUDAD', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 93025, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/prepare-dataset-from-yolo-format-to-coco-for-detr/34894/5', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hi. I would like to compare two nets using the same dataset, regardless being Transformer-based (DETR) vs Non-Transformer based (YOLOv5).
+I have already trained a model using Yolov5, such that my dataset is already split into train-val-test, in YOLO format. See Formatting table to visualize an example. My dataset folder looks like this:
.
+├── train
+ └── images
+ │ ├── ima1.png
+ │ ├── ima2.png
+ │ ├── ...
+ └── labels
+ │ ├── ima1.txt
+ │ ├── ima2.txt
+ │ ├── ...
+├── val
+ └── images
+ │ ├── ima3.png
+ │ ├── ima4.png
+ │ ├── ...
+ └── labels
+ │ ├── ima3.txt
+ │ ├── ima4.txt
+ │ ├── ...
+├── test
+ └── images
+ │ ├── ima5.png
+ │ ├── ima6.png
+ │ ├── ...
+ └── labels
+ │ ├── ima5.txt
+ │ ├── ima6.txt
+ │ ├── ...
+Now I want to convert it to COCO format. From Hugging Face documentation, DETR demands COCO format in labels, using JSON files. However, you are using a dataset loaded from Hugging Face datasets library. Moreover, I would like to know if I should create 3 JSON files, for each split, or 1 JSON file containing all. In the latter case, could you provide some documentation on how should the JSON file be defined?
+If there is any tutorial on how to prepare the data to feed DETR, based on my specs, it would be nice to post it here.
+Thank you for all!
In order to read it using load_dataset, it is a must to follow the same structure as defined
+here
I don’t know what happened here. For about 20-30 minutes the dataset card and data studio looked perfect and was working including the ability to query with SQL but now I have this error message and nothing works.
\nI was trying to add the metadata to my parquet file. It took several tries to get it right but maybe it was actually my 2nd to last try that was correct and the latest try is a disaster. Maybe I inadvertently over-wrote the good file.
\nCan anyone assist with debugging this and help me figure out how to restore the good file?
\nThe correct file should have the following columns:
\ncolumn 1 - year
\ncolumn 2 - path
\ncolumn 3 - file_name
\ncolumn 4 - record_number
\ncolumn 5 - nara_release_date
\ncolumn 6 - formerly_withheld
\ncolumn 7 - agency
\ncolumn 8 - document_date
\ncolumn 9 - document_type
\ncolumn 10 - file_number
\ncolumn 11 - to_name
\ncolumn 12 - from_name
\ncolumn 13 - title
\ncolumn 14 - number_of_pages
\ncolumn 15 - originator
\ncolumn 16 - record_series
\ncolumn 17 - review_date
\ncolumn 18 - comments
\ncolumn 19 - pages_released
\ncolumn 20 - content
The first file uploaded worked as well, it had only year, path, filename and content. These 16 new columns were inserted between filename and content.
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-05T14:55:06.888Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 19, 'reads': 6, 'readers_count': 5, 'score': 111.2, 'yours': False, 'topic_id': 153590, 'topic_slug': 'the-full-dataset-viewer-is-not-available-click-to-read-why-only-showing-a-preview-of-the-rows', 'display_username': 'Bill', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/datasets/mysocratesnote/jfk-files-text', 'internal': False, 'reflection': False, 'title': 'mysocratesnote/jfk-files-text · Datasets at Hugging Face', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91697, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-full-dataset-viewer-is-not-available-click-to-read-why-only-showing-a-preview-of-the-rows/153590/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 219935, 'name': 'Bill', 'username': 'mysocratesnote', 'avatar_template': '/user_avatar/discuss.huggingface.co/mysocratesnote/{size}/46167_2.png', 'created_at': '2025-05-05T19:11:08.441Z', 'cooked': 'Turns out uploading a .csv with a different number of columns even in a different directory broke it.
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-05T19:11:08.441Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 153590, 'topic_slug': 'the-full-dataset-viewer-is-not-available-click-to-read-why-only-showing-a-preview-of-the-rows', 'display_username': 'Bill', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91697, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/the-full-dataset-viewer-is-not-available-click-to-read-why-only-showing-a-preview-of-the-rows/153590/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 220026, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-06T07:11:25.083Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-06T07:11:25.083Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.8, 'yours': False, 'topic_id': 153590, 'topic_slug': 'the-full-dataset-viewer-is-not-available-click-to-read-why-only-showing-a-preview-of-the-rows', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/the-full-dataset-viewer-is-not-available-click-to-read-why-only-showing-a-preview-of-the-rows/153590/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I don’t know what happened here. For about 20-30 minutes the dataset card and data studio looked perfect and was working including the ability to query with SQL but now I have this error message and nothing works.
+I was trying to add the metadata to my parquet file. It took several tries to get it right but maybe it was actually my 2nd to last try that was correct and the latest try is a disaster. Maybe I inadvertently over-wrote the good file.
+Can anyone assist with debugging this and help me figure out how to restore the good file?
+The correct file should have the following columns:
+column 1 - year
+column 2 - path
+column 3 - file_name
+column 4 - record_number
+column 5 - nara_release_date
+column 6 - formerly_withheld
+column 7 - agency
+column 8 - document_date
+column 9 - document_type
+column 10 - file_number
+column 11 - to_name
+column 12 - from_name
+column 13 - title
+column 14 - number_of_pages
+column 15 - originator
+column 16 - record_series
+column 17 - review_date
+column 18 - comments
+column 19 - pages_released
+column 20 - content
The first file uploaded worked as well, it had only year, path, filename and content. These 16 new columns were inserted between filename and content.
",Turns out uploading a .csv with a different number of columns even in a different directory broke it.
+HF Playground Incorrect Billing -,https://discuss.huggingface.co/t/hf-playground-incorrect-billing/153328,153328,5,2025-05-03 12:01:35.655000+00:00,"[{'id': 219558, 'name': 'Kwabena Anim', 'username': 'KwabsHug', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/k/ba8739/{size}.png', 'created_at': '2025-05-03T12:01:35.766Z', 'cooked': 'Hello All, I was testing the HF playground and all my requests were only $0.20, I was testing in the window on the model page now my total is $9.08 (Model is Qwen/Qwen3-235B-A22B) Where can I find the HF Inference pricing and why is it so high? I got at best 10k tokens for price of Millions
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-03T12:11:46.503Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 23, 'reads': 8, 'readers_count': 7, 'score': 131.6, 'yours': False, 'topic_id': 153328, 'topic_slug': 'hf-playground-incorrect-billing', 'display_username': 'Kwabena Anim', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31391, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-playground-incorrect-billing/153328/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 219616, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-05-03T23:07:53.607Z', 'cooked': 'It seems that the criteria have changed. In other words, when using large models, the cost per request becomes expensive.
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-03T23:07:53.607Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 6, 'readers_count': 5, 'score': 16.2, 'yours': False, 'topic_id': 153328, 'topic_slug': 'hf-playground-incorrect-billing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/inference-api-cost-changed-for-meta-llama-3-3-70b/149074/3', 'internal': True, 'reflection': False, 'title': 'Inference API cost changed for meta-llama-3.3-70b?', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-playground-incorrect-billing/153328/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 219763, 'name': 'Andrew J tokar', 'username': 'Zelgodiz', 'avatar_template': '/user_avatar/discuss.huggingface.co/zelgodiz/{size}/45662_2.png', 'created_at': '2025-05-05T04:08:43.555Z', 'cooked': 'Starting in March, usage now takes into account compute time x price of the hardware
\n
It sounds like the pricing jumped unexpectedly! Hugging Face’s inference costs can vary based on the model’s size, provider, and token usage. The Qwen/Qwen3-235B-A22B model is a Mixture-of-Experts (MoE) model with 235 billion parameters, which means it can be significantly more expensive than smaller models43dcd9a7-70db-4a1f-b0ae-981daa162054.
\nYou can check Hugging Face’s official inference pricing on their model page or explore detailed cost breakdowns on LLM Stats.
\nIf you need help optimizing your usage, I can suggest ways to reduce token consumption!
Okay, so we are charged per time on HF inference API which means for now the solution is to use the other providers? Also is there a way to disable providers you dont want to use?
\nAlso is there a way to set a spending ceiling for my account?
\nIf I used R1 for the same task it wouldnt have cost this much through replicate for example.
The payment limit is set to $100 by default. (I think it was already there when I first joined HF.)
\nChanging this should be sufficient for personal use.
Detailed limits for the Inference API can apparently be set for Enterprise subscriptions. Well, if multiple people are using it, it’s more convenient to have separate limits for each service.
\nIndividual on/off settings for Inference Providers can be configured on the settings page.
\n\n\nEdit:
\n\n\nThe payment limit is set to $100 by default
\n
Oh… It was wrong…
\n', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-05-05T21:32:43.345Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 153328, 'topic_slug': 'hf-playground-incorrect-billing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/inference-api-budget-billing-limit/13239/14', 'internal': True, 'reflection': False, 'title': 'Inference API budget, billing limit', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/inference-api-budget-billing-limit/13239/13', 'internal': True, 'reflection': False, 'title': 'Inference API budget, billing limit', 'clicks': 0}, {'url': 'https://huggingface.co/docs/inference-providers/pricing', 'internal': False, 'reflection': False, 'title': 'Pricing and Billing', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hf-playground-incorrect-billing/153328/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 219939, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-05T19:28:48.453Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-05-05T19:28:48.453Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 153328, 'topic_slug': 'hf-playground-incorrect-billing', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/hf-playground-incorrect-billing/153328/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello All, I was testing the HF playground and all my requests were only $0.20, I was testing in the window on the model page now my total is $9.08 (Model is Qwen/Qwen3-235B-A22B) Where can I find the HF Inference pricing and why is it so high? I got at best 10k tokens for price of Millions
","The payment limit is set to $100 by default. (I think it was already there when I first joined HF.)
+Changing this should be sufficient for personal use.
Detailed limits for the Inference API can apparently be set for Enterprise subscriptions. Well, if multiple people are using it, it’s more convenient to have separate limits for each service.
+Individual on/off settings for Inference Providers can be configured on the settings page.
+ + +Edit:
+++The payment limit is set to $100 by default
+
Oh… It was wrong…
+" +Adding additional metadata columns to a .parque file from .xlsx files,https://discuss.huggingface.co/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017,152017,12,2025-04-23 18:50:05.289000+00:00,"[{'id': 217777, 'name': 'Bill', 'username': 'mysocratesnote', 'avatar_template': '/user_avatar/discuss.huggingface.co/mysocratesnote/{size}/46167_2.png', 'created_at': '2025-04-23T18:50:05.356Z', 'cooked': 'I just created a data set containing extracted text from the JFK Files.
\nEach release had an accompanying .xlsx file with a bunch of metadata including: Record Num, NARA Release Date, Formerly Withheld, Doc Date, Doc Type, Doc Type, File Num, To Name, From Name, Title, Num Pages, Originator, Record Series, Review Date, Comments, Pages Released
\nRecord Num - Record Number, also sometimes the filename less the extension but sometimes not.
\nNARA Release Date - Date archives(.)org released the file
\nFormerly Withheld - Reason for withholding the document
\nDoc Date - Original document date
\nDoc Type - Paper, audio tape, etc.
\nFile Num - File Number
\nTo Name - Who the document was addressed to
\nFrom Name - Who sent the document
\nTitle - Document title
\nNum Pages - Total number of pages in the document
\nOriginator - Where the document came from, often CIA or FBI
\nRecord Series - In this case they may all be ‘JFK’
\nReview Date - Date the document was reviewed for release
\nComments - Comments
\nPages Released - Number of pages released
It seems like the parque format is ideal to attach all this meta data to the content of the files and while this initially looks like a straight forward task, it’s a bit more challenging because:
\nThe same record number can refer to multiple files and a single file can have multiple record numbers.
\nSometimes the record number is the file name (less the extension), sometimes it’s a “dicid” (whatever that is) and sometimes the files follow no standard naming convention at all.
\nEach release has a different format for the .xlsx files.
\n2025 seems to have standardized on the record number for the file name and no .xlsx is provided. We only have filenames and NARA Release Date. But, many (maybe even all?) of these files were previously released (often with more redactions , blank or missing pages) and have meta data in the .xlsx files from previous releases.
\nMany of the same files appear again and again in subsequent releases usually with additional pages and/or less redactions.
\nThe 2017-2018 release is by far the largest and many files appear twice within the same release.
\nThis may be a trivial task for an experienced data scientist but it’s challenging for me therefore I’m reaching out to see if anyone can suggest the best approach.
', 'post_number': 1, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-24T05:52:21.958Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 20, 'reads': 3, 'readers_count': 2, 'score': 115.6, 'yours': False, 'topic_id': 152017, 'topic_slug': 'adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files', 'display_username': 'Bill', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 3, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/noops888/jfk-files-text/tree/main/downloader_scripts/xlsx', 'internal': False, 'reflection': False, 'title': 'jfk-files-text/downloader_scripts/xlsx at main · noops888/jfk-files-text · GitHub', 'clicks': 0}, {'url': 'https://huggingface.co/datasets/mysocratesnote/jfk-files-text', 'internal': False, 'reflection': False, 'title': 'mysocratesnote/jfk-files-text · Datasets at Hugging Face', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91697, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 217801, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-23T22:37:20.357Z', 'cooked': 'The xlsx format is often difficult to handle with software, so it would be better to convert it to csv (using Python or some kind of GUI tool) and then read it with the datasets library…
\nIncidentally, it will be converted to parquet format when it is read.
\nThe text is small, so size is not really an issue, but I think it would be better to copy it for multiple references. Is there a good way to convert complex xlsx files…?
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-23T22:37:20.357Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 3, 'readers_count': 2, 'score': 25.6, 'yours': False, 'topic_id': 152017, 'topic_slug': 'adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://www.geeksforgeeks.org/convert-excel-to-csv-in-python/', 'internal': False, 'reflection': False, 'title': 'Convert Excel to CSV in Python | GeeksforGeeks', 'clicks': 0}, {'url': 'https://huggingface.co/docs/datasets/en/loading', 'internal': False, 'reflection': False, 'title': 'Load', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 217962, 'name': 'Bill', 'username': 'mysocratesnote', 'avatar_template': '/user_avatar/discuss.huggingface.co/mysocratesnote/{size}/46167_2.png', 'created_at': '2025-04-24T15:59:19.655Z', 'cooked': 'Hi again @John6666 converting to .csv is no problem using python or just saving it to CSV from Exel - there are only four files. They are large but not super massive. The problem arises from a few different issues, inconsistent format of the spreadsheet. Record numbers that refer to multiple files but also single files that have multiple record numbers. Duplicate file listings in the spreadsheets (probably due to the record number issue), and some bad data:
\n34 files in the 2022 release and 5 files in the 2021 release tie to multiple record numbers listed in the .xlsx files which have more rows than unique file names (13,263 and 1,491 resptively). The 2017-2018 release xlsx filecontains 6 bad links, but the 2017-2018 release website lists two files not included in the xlsx in the /additional path. With two exceptions all .md files match up to .pdf files, the two exceptions match to .mp3 files.
\nnational-archives-jfk-assassination-records-2017-2018-release.xlsx (17 columns, 54,636 data rows, 1 header)
\nColumns: File Name, Record Num, NARA Release Date, Formerly Withheld, Agency, Doc Date, Doc Type. File Num\tTo Name, From Name, Title, Num Pages, Originator, Record Series, Review Date, Comments, Pages Released.
\nnational-archives-jfk-assassination-records-2021-release.xlsx (16 columns, 1,491 data rows, 1 header)
\nColumns: Record Number, File Title, NARA Release Date, Formerly Withheld, Document Date, Document Type, File Number., To, From, Title, Original Document Pages, Originator, Record Series, Review Date, Comments, Document Pages in PDF
\nFile Title is the same as File Name
\nDocument Pages in PDF is the same as Pages Released
\nAgency is missing (often the same as “Originator” but sometimes different).
national-archives-jfk-assassination-records-2022-release.xlsx (16 columns, 13,264 data rows, 1 header)
\nColumns: File Name, Record Num, NARA Release Date, Formerly Withheld, Doc Date, Doc Type, File Num\tTo Name, From Name,\tTitle, Num Pages, Originator, Record Series, Review Date, Comments, Pages Released
\nFormat looks the same as the first file but is missing “Agency”
\nnational-archives-jfk-assassination-records-2023-release.xlsx (17 columns, 2693 data rows, 1 header)
\nColumns: File Name, Record Num, NARA Release Date, Formerly Withheld, Agency, Doc Date, Doc Type\tFile Num, To Name, From Name, Title, Num Pages, Originator, Record Series, Review Date, Comments, Pages Released
\nBack to the first file’s format, Agency column is back but it’s blank for this release.
\n2025-release.xlsx (2 columns, 2,566 data rows, 1 header)
\nColumns: Record Number, NARA Release Date
\nThere was no .xlsx provided for 2025, this is the only available information from the website which mirrors the .xlsx for previous years.
\nFor an experienced developer I’m sure this is easy but I’m not sure how to go about because of all the inconsistencies and discrepancies. It’s not a simple 1:1 mapping. But, having all this metadata in the parque file and standardized as best as possible would definitely make for a much better data set.
\nIt would make sense to standardize on the column headings used in 3 out of the 4 files and to leave the columns blank where data wasn’t provided.
\nIf anyone can offer some advice on the best way to do this without introducing a bunch of data errors it would be much appreciated.
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-24T15:59:19.655Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 25.6, 'yours': False, 'topic_id': 152017, 'topic_slug': 'adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files', 'display_username': 'Bill', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://www.archives.gov/research/jfk/release-2017-2018', 'internal': False, 'reflection': False, 'title': 'JFK Assassination Records - 2017-2018 Additional Documents Release | National Archives', 'clicks': 0}, {'url': 'https://www.archives.gov/files/research/jfk/national-archives-jfk-assassination-records-2017-2018-release.xlsx', 'internal': False, 'reflection': False, 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91697, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218079, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-25T03:21:47.447Z', 'cooked': 'I’m not a data scientist, so this is just a general observation, but when dealing with text-based data, it’s easier for the computer to process if you align the data to the larger number.
\nRegardless of whether individual data points exist or not, it’s best to add all possible columns to all data.
And for complete irregularities like the mp3 part, it’s faster and more reliable to handle them manually. Just because you have the tools doesn’t mean you have to do it by hand—no one has decided that.
\nby Hugging Chat: HuggingChat
\nTo standardize the inconsistent spreadsheet data from the JFK assassination records releases, follow this structured approach:
\nRead and Load Data
\npandas library to read each Excel file into a DataFrame.import pandas as pd\n\nfiles = [\'2017-2018.xlsx\', \'2021.xlsx\', \'2022.xlsx\', \'2023.xlsx\', \'2025.xlsx\']\ndfs = []\nfor file in files:\n dfs.append(pd.read_excel(file))\nStandardize Column Names
\ncolumn_mapping = {\n \'File Name\': \'File Name\',\n \'Record Num\': \'Record Number\',\n \'NARA Release Date\': \'Release Date\',\n \'Formerly Withheld\': \'Withheld\',\n \'Agency\': \'Agency\',\n \'Doc Date\': \'Document Date\',\n \'Doc Type\': \'Document Type\',\n \'File Num To Name\': \'File Number\',\n \'From Name\': \'From\',\n \'Title\': \'Title\',\n \'Num Pages\': \'Pages\',\n \'Originator\': \'Originator\',\n \'Record Series\': \'Series\',\n \'Review Date\': \'Review Date\',\n \'Comments\': \'Comments\',\n \'Pages Released\': \'Released Pages\'\n}\nfor df in dfs:\n df.columns = [column_mapping.get(col, col) for col in df.columns]\nHandle Missing Columns
\nNaN where data is unavailable.all_columns = set()\nfor df in dfs:\n all_columns.update(df.columns)\nall_columns = list(all_columns)\n\nfor df in dfs:\n missing_cols = [col for col in all_columns if col not in df.columns]\n for col in missing_cols:\n df[col] = pd.NA\nIdentify and Remove Duplicates
\nfor df in dfs:\n df.drop_duplicates(subset=[\'Record Number\', \'File Name\'], inplace=True)\nClean Data and Validate Paths
\nimport os\n\nfor df in dfs:\n for index, row in df.iterrows():\n file_path = row[\'File Name\']\n if not os.path.exists(file_path):\n # Handle invalid path; log or flag\n print(f""Invalid path: {file_path}"")\n # Correct extension for known exceptions\n if file_path.endswith(\'.md\') and row[\'File Type\'] == \'mp3\':\n df.at[index, \'File Name\'] = file_path.replace(\'.md\', \'.mp3\')\nConcatenate DataFrames
\ncombined_df = pd.concat(dfs, ignore_index=True)\nSave to Standardized CSV
\ncombined_df.to_csv(\'standardized_jfk_records.csv\', index=False)\nThis approach systematically addresses the inconsistencies and discrepancies in the data, ensuring a standardized and clean dataset is produced.
', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-25T03:21:47.447Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 3, 'readers_count': 2, 'score': 30.6, 'yours': False, 'topic_id': 152017, 'topic_slug': 'adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/chat/', 'internal': False, 'reflection': False, 'title': 'HuggingChat', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 218099, 'name': 'Bill', 'username': 'mysocratesnote', 'avatar_template': '/user_avatar/discuss.huggingface.co/mysocratesnote/{size}/46167_2.png', 'created_at': '2025-04-25T06:39:46.293Z', 'cooked': 'That sounds like a very logical approach that will address all the issues, except the duplicate file listings which are multiple record numbers that apply to the same file. That needs to get into the final data. I guess the inverse were multiple files have the same record number would sort itself out automatically. You’re right mp3 and the few broken links can be handled manually.
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-25T06:39:46.293Z', 'reply_count': 0, 'reply_to_post_number': 4, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 4, 'readers_count': 3, 'score': 30.8, 'yours': False, 'topic_id': 152017, 'topic_slug': 'adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files', 'display_username': 'Bill', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91697, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 219883, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-05T14:32:31.129Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-05-05T14:32:31.129Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 2, 'readers_count': 1, 'score': 5.4, 'yours': False, 'topic_id': 152017, 'topic_slug': 'adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/adding-additional-metadata-columns-to-a-parque-file-from-xlsx-files/152017/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I just created a data set containing extracted text from the JFK Files.
+Each release had an accompanying .xlsx file with a bunch of metadata including: Record Num, NARA Release Date, Formerly Withheld, Doc Date, Doc Type, Doc Type, File Num, To Name, From Name, Title, Num Pages, Originator, Record Series, Review Date, Comments, Pages Released
+Record Num - Record Number, also sometimes the filename less the extension but sometimes not.
+NARA Release Date - Date archives(.)org released the file
+Formerly Withheld - Reason for withholding the document
+Doc Date - Original document date
+Doc Type - Paper, audio tape, etc.
+File Num - File Number
+To Name - Who the document was addressed to
+From Name - Who sent the document
+Title - Document title
+Num Pages - Total number of pages in the document
+Originator - Where the document came from, often CIA or FBI
+Record Series - In this case they may all be ‘JFK’
+Review Date - Date the document was reviewed for release
+Comments - Comments
+Pages Released - Number of pages released
It seems like the parque format is ideal to attach all this meta data to the content of the files and while this initially looks like a straight forward task, it’s a bit more challenging because:
+The same record number can refer to multiple files and a single file can have multiple record numbers.
+Sometimes the record number is the file name (less the extension), sometimes it’s a “dicid” (whatever that is) and sometimes the files follow no standard naming convention at all.
+Each release has a different format for the .xlsx files.
+2025 seems to have standardized on the record number for the file name and no .xlsx is provided. We only have filenames and NARA Release Date. But, many (maybe even all?) of these files were previously released (often with more redactions , blank or missing pages) and have meta data in the .xlsx files from previous releases.
+Many of the same files appear again and again in subsequent releases usually with additional pages and/or less redactions.
+The 2017-2018 release is by far the largest and many files appear twice within the same release.
+This may be a trivial task for an experienced data scientist but it’s challenging for me therefore I’m reaching out to see if anyone can suggest the best approach.
","I’m not a data scientist, so this is just a general observation, but when dealing with text-based data, it’s easier for the computer to process if you align the data to the larger number.
+Regardless of whether individual data points exist or not, it’s best to add all possible columns to all data.
And for complete irregularities like the mp3 part, it’s faster and more reliable to handle them manually. Just because you have the tools doesn’t mean you have to do it by hand—no one has decided that.
+by Hugging Chat: HuggingChat
+To standardize the inconsistent spreadsheet data from the JFK assassination records releases, follow this structured approach:
+Read and Load Data
+pandas library to read each Excel file into a DataFrame.import pandas as pd
+
+files = ['2017-2018.xlsx', '2021.xlsx', '2022.xlsx', '2023.xlsx', '2025.xlsx']
+dfs = []
+for file in files:
+ dfs.append(pd.read_excel(file))
+Standardize Column Names
+column_mapping = {
+ 'File Name': 'File Name',
+ 'Record Num': 'Record Number',
+ 'NARA Release Date': 'Release Date',
+ 'Formerly Withheld': 'Withheld',
+ 'Agency': 'Agency',
+ 'Doc Date': 'Document Date',
+ 'Doc Type': 'Document Type',
+ 'File Num To Name': 'File Number',
+ 'From Name': 'From',
+ 'Title': 'Title',
+ 'Num Pages': 'Pages',
+ 'Originator': 'Originator',
+ 'Record Series': 'Series',
+ 'Review Date': 'Review Date',
+ 'Comments': 'Comments',
+ 'Pages Released': 'Released Pages'
+}
+for df in dfs:
+ df.columns = [column_mapping.get(col, col) for col in df.columns]
+Handle Missing Columns
+NaN where data is unavailable.all_columns = set()
+for df in dfs:
+ all_columns.update(df.columns)
+all_columns = list(all_columns)
+
+for df in dfs:
+ missing_cols = [col for col in all_columns if col not in df.columns]
+ for col in missing_cols:
+ df[col] = pd.NA
+Identify and Remove Duplicates
+for df in dfs:
+ df.drop_duplicates(subset=['Record Number', 'File Name'], inplace=True)
+Clean Data and Validate Paths
+import os
+
+for df in dfs:
+ for index, row in df.iterrows():
+ file_path = row['File Name']
+ if not os.path.exists(file_path):
+ # Handle invalid path; log or flag
+ print(f""Invalid path: {file_path}"")
+ # Correct extension for known exceptions
+ if file_path.endswith('.md') and row['File Type'] == 'mp3':
+ df.at[index, 'File Name'] = file_path.replace('.md', '.mp3')
+Concatenate DataFrames
+combined_df = pd.concat(dfs, ignore_index=True)
+Save to Standardized CSV
+combined_df.to_csv('standardized_jfk_records.csv', index=False)
+This approach systematically addresses the inconsistencies and discrepancies in the data, ensuring a standardized and clean dataset is produced.
" +Why `inv_freq` when computing frequencies for RoPE,https://discuss.huggingface.co/t/why-inv-freq-when-computing-frequencies-for-rope/153106,153106,9,2025-05-01 09:58:34.624000+00:00,"[{'id': 219283, 'name': 'Ye Zhiling', 'username': 'yzlnew', 'avatar_template': '/user_avatar/discuss.huggingface.co/yzlnew/{size}/46705_2.png', 'created_at': '2025-05-01T09:58:34.687Z', 'cooked': 'I’m getting confused at the naming here,
\n # Compute the inverse frequencies\n inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.int64).to(device=device, dtype=torch.float) / dim))\n return inv_freq, attention_factor\nThis inv_freq is actually meaning frequencies for each dimension for RoPE. What does inv mean here?
Reply to yzlnew on ‘Why inv_freq when computing frequencies for RoPE’
Hi @yzlnew! Great question — this is a common source of confusion when diving into RoPE implementation details. Let me break it down clearly:
\ninv_freq in the context of RoPE?In most implementations of Rotary Positional Embeddings (RoPE), the inv_freq refers to the inverse frequency used to compute the positional encodings for each embedding dimension. It’s inspired by the same idea behind sinusoidal embeddings in the original Transformer paper, where different dimensions of the input are assigned sinusoidal functions with different wavelengths.
The key lies in this line:
\ninv_freq = 1.0 / (base ** (torch.arange(0, dim, 2) / dim))\nThis gives you a vector of inverse frequencies — meaning higher frequency values (shorter wavelengths) for lower dimensions, and lower frequency values (longer wavelengths) for higher dimensions.
\nSo for example:
\ndim=0, you might have an inv_freq like 1/10000^0 = 1dim=2, you get 1/10000^(2/dim), and so on…This mirrors the logarithmic spacing of frequencies, enabling smooth interpolation and generalization across positions.
\nThen, when you later multiply position_ids * inv_freq, you get a phase angle for each position, which is passed to sin() and cos() to rotate the query/key vectors — hence the term “rotary”.
inv_freq = inverse frequency per dimensionLet me know if you’d like a visual intuition or derivation behind the rotational aspect of RoPE — happy to elaborate!
\nCheers,
\nHaruthai AI (Sunny)
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-03T01:22:58.384Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 153106, 'topic_slug': 'why-inv-freq-when-computing-frequencies-for-rope', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/why-inv-freq-when-computing-frequencies-for-rope/153106/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m getting confused at the naming here,
+ # Compute the inverse frequencies
+ inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.int64).to(device=device, dtype=torch.float) / dim))
+ return inv_freq, attention_factor
+This inv_freq is actually meaning frequencies for each dimension for RoPE. What does inv mean here?
Reply to yzlnew on ‘Why inv_freq when computing frequencies for RoPE’
Hi @yzlnew! Great question — this is a common source of confusion when diving into RoPE implementation details. Let me break it down clearly:
+inv_freq in the context of RoPE?In most implementations of Rotary Positional Embeddings (RoPE), the inv_freq refers to the inverse frequency used to compute the positional encodings for each embedding dimension. It’s inspired by the same idea behind sinusoidal embeddings in the original Transformer paper, where different dimensions of the input are assigned sinusoidal functions with different wavelengths.
The key lies in this line:
+inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2) / dim))
+This gives you a vector of inverse frequencies — meaning higher frequency values (shorter wavelengths) for lower dimensions, and lower frequency values (longer wavelengths) for higher dimensions.
+So for example:
+dim=0, you might have an inv_freq like 1/10000^0 = 1dim=2, you get 1/10000^(2/dim), and so on…This mirrors the logarithmic spacing of frequencies, enabling smooth interpolation and generalization across positions.
+Then, when you later multiply position_ids * inv_freq, you get a phase angle for each position, which is passed to sin() and cos() to rotate the query/key vectors — hence the term “rotary”.
inv_freq = inverse frequency per dimensionLet me know if you’d like a visual intuition or derivation behind the rotational aspect of RoPE — happy to elaborate!
+Cheers,
+Haruthai AI (Sunny)
I am using smolagents library with HfAPIModel. Where can I find the pricing related to the models I can use with it? Do I pay based on tokens or amount of requests?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-04-30T10:39:47.855Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 57, 'reads': 7, 'readers_count': 6, 'score': 301.4, 'yours': False, 'topic_id': 153001, 'topic_slug': 'hfapimodel-pricing', 'display_username': 'Giuseppe Boezio', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 89270, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hfapimodel-pricing/153001/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 219174, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-30T12:10:12.190Z', 'cooked': '\n\nProbably the number of requests multiplied by the price of the GPU used for that model. For exact details, please consult Hugging Face. billing@huggingface.co
', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-05-01T15:19:55.354Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 7, 'readers_count': 6, 'score': 16.4, 'yours': False, 'topic_id': 153001, 'topic_slug': 'hfapimodel-pricing', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/inference-providers/en/pricing#hf-inference-cost', 'internal': False, 'reflection': False, 'title': 'Pricing and Billing', 'clicks': 5}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/hfapimodel-pricing/153001/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 219404, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-05-02T08:00:24.283Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-05-02T08:00:24.283Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 153001, 'topic_slug': 'hfapimodel-pricing', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/hfapimodel-pricing/153001/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]",I am using smolagents library with HfAPIModel. Where can I find the pricing related to the models I can use with it? Do I pay based on tokens or amount of requests?
," ++Probably the number of requests multiplied by the price of the GPU used for that model. For exact details, please consult Hugging Face. billing@huggingface.co
" +Server-side problems,https://discuss.huggingface.co/t/server-side-problems/150852,150852,24,2025-04-16 15:40:07.811000+00:00,"[{'id': 216187, 'name': 'Edward J. Schwartz', 'username': 'ejschwartz', 'avatar_template': '/user_avatar/discuss.huggingface.co/ejschwartz/{size}/16902_2.png', 'created_at': '2025-04-16T15:40:07.883Z', 'cooked': 'I’ve encountered two strange errors in a short period of time.
\nSpace: Aidapal Space - a Hugging Face Space by ejschwartz
\nI created a new space. I committed app.py and pushed, and got an error that was roughly “Unable to find app.py” in the runtime logs.
I just added and committed requirements.txt and received the following build error.
\n\nBoth problems seem to be related to not finding a recently committed file. Manually doing a factory rebuild seems to mitigate the problem.
', 'post_number': 1, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-04-16T15:40:36.169Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 64, 'reads': 11, 'readers_count': 10, 'score': 332.2, 'yours': False, 'topic_id': 150852, 'topic_slug': 'server-side-problems', 'display_username': 'Edward J. Schwartz', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/spaces/ejschwartz/aidapal-space', 'internal': False, 'reflection': False, 'title': 'Aidapal Space - a Hugging Face Space by ejschwartz', 'clicks': 3}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 22191, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/server-side-problems/150852/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 216259, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-17T03:39:05.812Z', 'cooked': 'It might be the same rollback bug that occurred in Dev mode before.
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-04-17T03:39:05.812Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 1.6, 'yours': False, 'topic_id': 150852, 'topic_slug': 'server-side-problems', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/hugging-face-space-keeps-using-an-old-commit-despite-redeploys/139695/4', 'internal': True, 'reflection': False, 'title': 'Hugging Face Space Keeps Using an Old Commit Despite Redeploys', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/server-side-problems/150852/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216348, 'name': 'Edward J. Schwartz', 'username': 'ejschwartz', 'avatar_template': '/user_avatar/discuss.huggingface.co/ejschwartz/{size}/16902_2.png', 'created_at': '2025-04-17T13:01:20.623Z', 'cooked': 'I was not using DEV mode. I’ll report if I run into any more problems today.
Whether it will be fixed or not, it’s an unknown issue…
\nIt seems that it’s OK to report the hub issue below.
\n', 'post_number': 4, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-04-17T13:07:58.375Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 8, 'readers_count': 7, 'score': 16.6, 'yours': False, 'topic_id': 150852, 'topic_slug': 'server-side-problems', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/hub-docs/issues', 'internal': False, 'reflection': False, 'title': 'GitHub · Where software is built', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/server-side-problems/150852/4', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216374, 'name': 'Edward J. Schwartz', 'username': 'ejschwartz', 'avatar_template': '/user_avatar/discuss.huggingface.co/ejschwartz/{size}/16902_2.png', 'created_at': '2025-04-17T15:33:13.286Z', 'cooked': '\n\nStill an issue.
\n\nHere the space fails to parse a JSON file that is committed to the repository.
\nI will report to HF.
\n
Disregard this message This was my mistake. The file I was loading was jsonl but was labeled as json. I have not seen any problems since yesterday.
', 'post_number': 5, 'post_type': 1, 'posts_count': 7, 'updated_at': '2025-04-17T15:46:36.942Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 7, 'readers_count': 6, 'score': 36.4, 'yours': False, 'topic_id': 150852, 'topic_slug': 'server-side-problems', 'display_username': 'Edward J. Schwartz', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 22191, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/server-side-problems/150852/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 216383, 'name': 'Megan Riley', 'username': 'meganariley', 'avatar_template': '/user_avatar/discuss.huggingface.co/meganariley/{size}/20596_2.png', 'created_at': '2025-04-17T16:35:54.198Z', 'cooked': 'Hi! I’m glad to hear the issue is now resolved
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 8, 'post_type': 3, 'posts_count': 7, 'updated_at': '2025-05-01T13:46:17.194Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 1, 'readers_count': 0, 'score': 10.2, 'yours': False, 'topic_id': 150852, 'topic_slug': 'server-side-problems', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/server-side-problems/150852/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’ve encountered two strange errors in a short period of time.
+Space: Aidapal Space - a Hugging Face Space by ejschwartz
+I created a new space. I committed app.py and pushed, and got an error that was roughly “Unable to find app.py” in the runtime logs.
I just added and committed requirements.txt and received the following build error.
+ +Both problems seem to be related to not finding a recently committed file. Manually doing a factory rebuild seems to mitigate the problem.
","++Still an issue.
+ +Here the space fails to parse a JSON file that is committed to the repository.
+I will report to HF.
+
Disregard this message This was my mistake. The file I was loading was jsonl but was labeled as json. I have not seen any problems since yesterday.
" +Can the T5 model classify codes such as codebert-small-v1?,https://discuss.huggingface.co/t/can-the-t5-model-classify-codes-such-as-codebert-small-v1/152496,152496,5,2025-04-27 10:03:32.978000+00:00,"[{'id': 218451, 'name': 'Franck da COSTA', 'username': 'kirilinko', 'avatar_template': '/user_avatar/discuss.huggingface.co/kirilinko/{size}/46423_2.png', 'created_at': '2025-04-27T10:03:33.036Z', 'cooked': 'Hello.
\nI’m doing code classification with codebert-small-v1, but as the maximum sequence is 512 tokens, this may limit me when faced with a certain amount of code (because of the size). On the other hand, I’ve noticed that T5 has a greater margin as regards the maximum sequence. Is it possible to use the T5 model for sort code classification to have the same output as codebert-small-v1? In the sense that I have the probability of appearance of each class of vulnerability in the code?
I’m not familiar with it, but it seems possible.
\n\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-27T10:27:35.969Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 152496, 'topic_slug': 'can-the-t5-model-classify-codes-such-as-codebert-small-v1', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/Salesforce/codet5-base', 'internal': False, 'reflection': False, 'title': 'Salesforce/codet5-base · Hugging Face', 'clicks': 3}, {'url': 'https://arxiv.org/abs/2408.07181', 'internal': False, 'reflection': False, 'title': '[2408.07181] VulCatch: Enhancing Binary Vulnerability Detection through CodeT5 Decompilation and KAN Advanced Feature Extraction', 'clicks': 0}, {'url': 'https://huggingface.co/huggingface/CodeBERTa-small-v1', 'internal': False, 'reflection': False, 'title': 'huggingface/CodeBERTa-small-v1 · Hugging Face', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-the-t5-model-classify-codes-such-as-codebert-small-v1/152496/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218616, 'name': 'Franck da COSTA', 'username': 'kirilinko', 'avatar_template': '/user_avatar/discuss.huggingface.co/kirilinko/{size}/46423_2.png', 'created_at': '2025-04-28T09:12:37.985Z', 'cooked': 'But I’m a bit surprised, when I try to classify with “TFAutoModelForSequenceClassification”, I get an error telling me that model T5 is not compatible. However, with codeBert small, no problem. I want to try another model because, I lack performance in predictions. My current model manages to classify the code well according to the CWE around 8 classes, but not when the code is vulnerable (only two classes) Do you have any idea what to do?
', 'post_number': 3, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-28T09:16:37.704Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 4, 'readers_count': 3, 'score': 20.8, 'yours': False, 'topic_id': 152496, 'topic_slug': 'can-the-t5-model-classify-codes-such-as-codebert-small-v1', 'display_username': 'Franck da COSTA', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90907, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-the-t5-model-classify-codes-such-as-codebert-small-v1/152496/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218690, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-28T12:50:13.942Z', 'cooked': 'Hmm…
\n\n\n', 'post_number': 4, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-28T12:50:13.942Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 3, 'readers_count': 2, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 152496, 'topic_slug': 'can-the-t5-model-classify-codes-such-as-codebert-small-v1', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/huggingface/transformers/issues/10405', 'internal': False, 'reflection': False, 'title': 'Problem running T5 (configuration) with text classification · Issue #10405 · huggingface/transformers · GitHub', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-the-t5-model-classify-codes-such-as-codebert-small-v1/152496/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 219173, 'name': 'Franck da COSTA', 'username': 'kirilinko', 'avatar_template': '/user_avatar/discuss.huggingface.co/kirilinko/{size}/46423_2.png', 'created_at': '2025-04-30T11:23:13.244Z', 'cooked': 'even though T5 can be used very well for text-classification it remains a text-to-text only model. So you can only load the model via
\n
\nfrom transformers import AutoModelForConditionalGeneration
\nmodel = AutoModelForConditionalGeneration.from_pretrained(“t5-small”)
thank you !
', 'post_number': 5, 'post_type': 1, 'posts_count': 6, 'updated_at': '2025-04-30T11:23:13.244Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 15.4, 'yours': False, 'topic_id': 152496, 'topic_slug': 'can-the-t5-model-classify-codes-such-as-codebert-small-v1', 'display_username': 'Franck da COSTA', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 90907, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/can-the-t5-model-classify-codes-such-as-codebert-small-v1/152496/5', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 219233, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-04-30T23:24:02.666Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 6, 'post_type': 3, 'posts_count': 6, 'updated_at': '2025-04-30T23:24:02.666Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 1, 'readers_count': 0, 'score': 5.2, 'yours': False, 'topic_id': 152496, 'topic_slug': 'can-the-t5-model-classify-codes-such-as-codebert-small-v1', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/can-the-t5-model-classify-codes-such-as-codebert-small-v1/152496/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","Hello.
+I’m doing code classification with codebert-small-v1, but as the maximum sequence is 512 tokens, this may limit me when faced with a certain amount of code (because of the size). On the other hand, I’ve noticed that T5 has a greater margin as regards the maximum sequence. Is it possible to use the T5 model for sort code classification to have the same output as codebert-small-v1? In the sense that I have the probability of appearance of each class of vulnerability in the code?
Hmm…
+ ++" +Docling image captioning best VLM,https://discuss.huggingface.co/t/docling-image-captioning-best-vlm/152311,152311,13,2025-04-25 14:37:54.184000+00:00,"[{'id': 218203, 'name': 'Sean Bayly', 'username': 'swtb', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/8c91f0/{size}.png', 'created_at': '2025-04-25T14:37:54.254Z', 'cooked': 'even though T5 can be used very well for text-classification it remains a text-to-text only model. So you can only load the model via
+
+from transformers import AutoModelForConditionalGeneration
+model = AutoModelForConditionalGeneration.from_pretrained(“t5-small”)
What is the current SOTA model for captioning images in documents?
\nI need good descriptions of diagrams. Most of the ones I have seen have very basic descriptions “the image contains a woman in a blue dress”. I need more like “The figure shows a flowchart representing a process of… that starts with…and ends with…key steps are…”
\nOr “The image depicts a scene in which people walk about in a modern cafe, key elements of the cafes design are…”
\nIn other words I need a good paragraph that offers some insight into the image.
\nAny suggestions on models?
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-04-25T14:37:54.254Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 202, 'reads': 5, 'readers_count': 4, 'score': 1006.0, 'yours': False, 'topic_id': 152311, 'topic_slug': 'docling-image-captioning-best-vlm', 'display_username': 'Sean Bayly', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 37927, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/docling-image-captioning-best-vlm/152311/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 218212, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-25T15:33:04.696Z', 'cooked': 'I’m not sure which VLM is strong in understanding the context of image content…
\nHow about trying out some VLM that seem to perform well to some extent…
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-04-29T19:34:51.185Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 3, 'readers_count': 2, 'score': 15.6, 'yours': False, 'topic_id': 152311, 'topic_slug': 'docling-image-captioning-best-vlm', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/docling-image-captioning-best-vlm/152311/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","What is the current SOTA model for captioning images in documents?
+I need good descriptions of diagrams. Most of the ones I have seen have very basic descriptions “the image contains a woman in a blue dress”. I need more like “The figure shows a flowchart representing a process of… that starts with…and ends with…key steps are…”
+Or “The image depicts a scene in which people walk about in a modern cafe, key elements of the cafes design are…”
+In other words I need a good paragraph that offers some insight into the image.
+Any suggestions on models?
","I’m not sure which VLM is strong in understanding the context of image content…
+How about trying out some VLM that seem to perform well to some extent…
I encountered an issue where the character’s head is not fully displayed when generating images with IPAdapter. How can I resolve this problem? Below is a screenshot of my workflow.
\n
Hmm, I’m not familiar with ComfyUI…
\n', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-04-28T08:47:44.128Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 4, 'readers_count': 3, 'score': 0.6000000000000001, 'yours': False, 'topic_id': 152581, 'topic_slug': 'incomplete-character-head-display-when-using-ipadapter', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://github.com/cubiq/ComfyUI_IPAdapter_plus/issues/406', 'internal': False, 'reflection': False, 'title': 'IPAdapterTiled crops images with 4:5 AR · Issue #406 · cubiq/ComfyUI_IPAdapter_plus · GitHub', 'clicks': 2}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/incomplete-character-head-display-when-using-ipadapter/152581/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 218744, 'name': 'retrooisa', 'username': 'jamoce', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/j/96bed5/{size}.png', 'created_at': '2025-04-28T17:31:21.857Z', 'cooked': 'You’re definitely not alone – I’ve run into the same issue when using IPAdapter. It’s usually something to do with the scaling settings or the way the input image is being processed. Bit of tweaking usually sorts it! By the way, if you’re after solid help with this sort of thing, having real expertise in modern tech makes a huge difference. The Frontend Company, for example, specialises in cutting-edge frameworks like React, Angular, and Vue.js. You might find their hire frontend developer guide quite useful too.
', 'post_number': 3, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-05-01T15:20:25.350Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 4, 'readers_count': 3, 'score': 30.6, 'yours': False, 'topic_id': 152581, 'topic_slug': 'incomplete-character-head-display-when-using-ipadapter', 'display_username': 'retrooisa', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 92232, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/incomplete-character-head-display-when-using-ipadapter/152581/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}, {'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218856, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-04-29T05:32:14.562Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-04-29T05:32:14.562Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.2, 'yours': False, 'topic_id': 152581, 'topic_slug': 'incomplete-character-head-display-when-using-ipadapter', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/incomplete-character-head-display-when-using-ipadapter/152581/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I encountered an issue where the character’s head is not fully displayed when generating images with IPAdapter. How can I resolve this problem? Below is a screenshot of my workflow.
+
Hmm, I’m not familiar with ComfyUI…
+" +Colab cannot find HuggingFace dataset,https://discuss.huggingface.co/t/colab-cannot-find-huggingface-dataset/63448,63448,10,2023-11-24 21:18:42.821000+00:00,"[{'id': 100772, 'name': 'Seyyed Mohammad Moosavi', 'username': 'lnxdx', 'avatar_template': '/user_avatar/discuss.huggingface.co/lnxdx/{size}/20601_2.png', 'created_at': '2023-11-24T21:18:42.886Z', 'cooked': 'When I try to run the following code to load a dataset from Hugging Face hub to google Colab, I get an error!
\n! pip install transformers datasets\nfrom datasets import load_dataset\ncv_13 = load_dataset(""mozilla-foundation/common_voice_13_0"", ""en"", split=""train"")\n<ipython-input-9-4d772f75be89> in <cell line: 3>()\n 1 from datasets import load_dataset\n 2 \n----> 3 cv_13 = load_dataset(""mozilla-foundation/common_voice_13_0"", ""en"", split=""train"")\n\n2 frames\n/usr/local/lib/python3.10/dist-packages/datasets/load.py in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, **download_kwargs)\n 1505 raise e1 from None\n 1506 if isinstance(e1, FileNotFoundError):\n-> 1507 raise FileNotFoundError(\n 1508 f""Couldn\'t find a dataset script at {relative_to_absolute_path(combined_path)} or any data file in the same directory. ""\n 1509 f""Couldn\'t find \'{path}\' on the Hugging Face Hub either: {type(e1).__name__}: {e1}""\n\nFileNotFoundError: Couldn\'t find a dataset script at /content/mozilla-foundation/common_voice_13_0/common_voice_13_0.py or any data file in the same directory. Couldn\'t find \'mozilla-foundation/common_voice_13_0\' on the Hugging Face Hub either: FileNotFoundError: Dataset \'mozilla-foundation/common_voice_13_0\' doesn\'t exist on the Hub. If the repo is private or gated, make sure to log in with `huggingface-cli login`.\nThe dataset exists in Huggingface hub and loads successfully in my local Jupiter Lab. What should I do?
', 'post_number': 1, 'post_type': 1, 'posts_count': 8, 'updated_at': '2023-11-24T21:18:42.886Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4822, 'reads': 145, 'readers_count': 144, 'score': 24003.8, 'yours': False, 'topic_id': 63448, 'topic_slug': 'colab-cannot-find-huggingface-dataset', 'display_username': 'Seyyed Mohammad Moosavi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/error-in-downloading-private-dataset/125836/4', 'internal': True, 'reflection': True, 'title': 'Error in downloading private dataset', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31952, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/colab-cannot-find-huggingface-dataset/63448/1', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 101062, 'name': 'Julien Chaumond', 'username': 'julien-c', 'avatar_template': '/user_avatar/discuss.huggingface.co/julien-c/{size}/41937_2.png', 'created_at': '2023-11-27T09:11:00.608Z', 'cooked': 'Which version of datasets are you using?
\ncc @lhoestq just in case
', 'post_number': 2, 'post_type': 1, 'posts_count': 8, 'updated_at': '2023-11-27T09:11:00.608Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 60, 'reads': 113, 'readers_count': 112, 'score': 342.4, 'yours': False, 'topic_id': 63448, 'topic_slug': 'colab-cannot-find-huggingface-dataset', 'display_username': 'Julien Chaumond', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': True, 'admin': True, 'staff': True, 'user_id': 4, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/colab-cannot-find-huggingface-dataset/63448/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 101084, 'name': 'Quentin Lhoest', 'username': 'lhoestq', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png', 'created_at': '2023-11-27T10:00:37.033Z', 'cooked': 'The Common Voice dataset is a gated dataset, so you need to log in to access it.
\nCan you try to log in using huggingface-cli login or pass
\na HF token load_dataset(..., token=...) ?
I logged in using huggingface-cli login and the dataset is currently being downloaded.
\ndatasets version is datasets-2.15.0-py3-none-any.whl.
I logged in using huggingface-cli login and the dataset is currently being downloaded. Thank you!
', 'post_number': 6, 'post_type': 1, 'posts_count': 8, 'updated_at': '2023-11-27T10:44:07.463Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 12, 'reads': 96, 'readers_count': 95, 'score': 79.0, 'yours': False, 'topic_id': 63448, 'topic_slug': 'colab-cannot-find-huggingface-dataset', 'display_username': 'Seyyed Mohammad Moosavi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76, 'username': 'lhoestq', 'name': 'Quentin Lhoest', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 31952, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/colab-cannot-find-huggingface-dataset/63448/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 135815, 'name': 'wangguan', 'username': 'wangguan1995', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/w/4bbf92/{size}.png', 'created_at': '2024-06-06T06:55:27.624Z', 'cooked': '#Dataset xxx doesn’t exist on the Hub or cannot be accessed
\nMeet similar problem can load public dataset, not for private dataset
I tried the same things. It does not work. Mine is a private dataset.
', 'post_number': 8, 'post_type': 1, 'posts_count': 8, 'updated_at': '2024-06-06T06:57:47.172Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 50, 'readers_count': 49, 'score': 30.0, 'yours': False, 'topic_id': 63448, 'topic_slug': 'colab-cannot-find-huggingface-dataset', 'display_username': 'wangguan', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76, 'username': 'lhoestq', 'name': 'Quentin Lhoest', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png'}, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52954, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/colab-cannot-find-huggingface-dataset/63448/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218634, 'name': 'yoldas', 'username': 'elifyoldas', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/e/bbce88/{size}.png', 'created_at': '2025-04-28T10:36:14.918Z', 'cooked': 'it works, thank you
', 'post_number': 9, 'post_type': 1, 'posts_count': 8, 'updated_at': '2025-04-28T10:36:14.918Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 11, 'readers_count': 10, 'score': 27.2, 'yours': False, 'topic_id': 63448, 'topic_slug': 'colab-cannot-find-huggingface-dataset', 'display_username': 'yoldas', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 76, 'username': 'lhoestq', 'name': 'Quentin Lhoest', 'avatar_template': '/user_avatar/discuss.huggingface.co/lhoestq/{size}/52888_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 92190, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/colab-cannot-find-huggingface-dataset/63448/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","When I try to run the following code to load a dataset from Hugging Face hub to google Colab, I get an error!
+! pip install transformers datasets
+from datasets import load_dataset
+cv_13 = load_dataset(""mozilla-foundation/common_voice_13_0"", ""en"", split=""train"")
+<ipython-input-9-4d772f75be89> in <cell line: 3>()
+ 1 from datasets import load_dataset
+ 2
+----> 3 cv_13 = load_dataset(""mozilla-foundation/common_voice_13_0"", ""en"", split=""train"")
+
+2 frames
+/usr/local/lib/python3.10/dist-packages/datasets/load.py in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, **download_kwargs)
+ 1505 raise e1 from None
+ 1506 if isinstance(e1, FileNotFoundError):
+-> 1507 raise FileNotFoundError(
+ 1508 f""Couldn't find a dataset script at {relative_to_absolute_path(combined_path)} or any data file in the same directory. ""
+ 1509 f""Couldn't find '{path}' on the Hugging Face Hub either: {type(e1).__name__}: {e1}""
+
+FileNotFoundError: Couldn't find a dataset script at /content/mozilla-foundation/common_voice_13_0/common_voice_13_0.py or any data file in the same directory. Couldn't find 'mozilla-foundation/common_voice_13_0' on the Hugging Face Hub either: FileNotFoundError: Dataset 'mozilla-foundation/common_voice_13_0' doesn't exist on the Hub. If the repo is private or gated, make sure to log in with `huggingface-cli login`.
+The dataset exists in Huggingface hub and loads successfully in my local Jupiter Lab. What should I do?
","The Common Voice dataset is a gated dataset, so you need to log in to access it.
+Can you try to log in using huggingface-cli login or pass
+a HF token load_dataset(..., token=...) ?
I have a question about how to specify arguments of custom TrainerCallback function. I read from some examples (e.g., doc) that users can specify custom arguments like model in the EmbeddingPlotCallback.on_evaluate(...) function. Here, model is not a predefined argument of the super class function TrainerCallback.on_evaluate(...) (doc).
I am wondering how the model is passed to this on_evaluate(...). Should I modify the Trainer class to make it call on_evaluate(...) with additional inputs? Or does the Trainer class handle additional arguments automatically? I have not yet found any examples about these. Any advice or points to relevant code sections/examples will be very helpful.
To supplement this inquiry with my motivation, I am experimenting with DPOTrainer while enabling synchronization of reference model, and I would like to log info about both the policy model and reference model. So, probably the inputs to the logging function would require two custom inputs for those two models. I think I can define two more arguments to my custom logging function, but I am not sure how I could pass the two models to my function.
\nAny comments will be greatly appreciated!
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-18T03:09:20.685Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 108, 'reads': 9, 'readers_count': 8, 'score': 536.8, 'yours': False, 'topic_id': 151063, 'topic_slug': 'how-to-write-custom-trainercallback-functions-with-custom-arguments', 'display_username': 'TTTTTC', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/setfit/main/how_to/callbacks', 'internal': False, 'reflection': False, 'title': 'Callbacks', 'clicks': 1}, {'url': 'https://huggingface.co/docs/transformers/main_classes/callback#transformers.TrainerCallback.on_evaluate', 'internal': False, 'reflection': False, 'title': 'Callbacks', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91116, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-write-custom-trainercallback-functions-with-custom-arguments/151063/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 216473, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-18T05:24:40.346Z', 'cooked': 'It seems to be automatic, or rather semi-manual. It seems that the trainer does not need to be modified.
\nby Hugging Chat: HuggingChat
\nTo specify additional arguments for a custom TrainerCallback function in the Hugging Face Transformers library, you need to design your callback class to accept and store these arguments during initialization. These arguments can then be accessed within the callback’s methods (e.g., on_evaluate). Below is a detailed explanation of how to achieve this:
TrainerCallback.__init__ method, accept any additional arguments you need (e.g., policy_model and ref_model).from transformers import TrainerCallback\n\nclass MyCustomCallback(TrainerCallback):\n def __init__(self, policy_model, ref_model):\n super().__init__()\n self.policy_model = policy_model\n self.ref_model = ref_model\n\n def on_evaluate(self, args, state, control, **kwargs):\n # Access your additional arguments here\n print(f""Evaluating with Policy Model: {self.policy_model}"")\n print(f""Evaluating with Reference Model: {self.ref_model}"")\n return super().on_evaluate(args, state, control, **kwargs)\nTrainer instance using add_callback().from transformers import Trainer\n\n# Initialize your models and trainer\ntrainer = Trainer(...) # Your existing trainer setup\n\n# Create an instance of your custom callback\ncallback = MyCustomCallback(policy_model=policy_model, ref_model=ref_model)\n\n# Add the callback to the trainer\ntrainer.add_callback(callback)\n\n# Start training\ntrainer.train()\ndef on_evaluate(self, args, state, control, **kwargs):\n # Use self.policy_model and self.ref_model here\n print(f""Policy Model: {self.policy_model}"")\n print(f""Reference Model: {self.ref_model}"")\n return super().on_evaluate(args, state, control, **kwargs)\nTrainer class in Hugging Face Transformers does not automatically pass additional arguments to callbacks. Instead, you must design your callback to accept and store these arguments during initialization.__init__ method in your custom callback, you can pass any arguments (e.g., policy_model and ref_model) and store them for use within the callback’s methods.Trainer class will call the callback methods (e.g., on_evaluate) during training, and your custom callback can access the stored arguments to perform the desired functionality.Trainer class supports custom callbacks by allowing users to add them via add_callback(). The Trainer does not restrict the number or type of arguments passed to the callback during initialization.on_evaluate, on_train_begin, etc.).Do I need to modify the Trainer class?
Trainer class already supports custom callbacks through the add_callback() method. You do not need to modify it to pass additional arguments to your callback.Can I pass any type of argument to the callback?
\nWhere should I store the additional arguments?
\n__init__ method of your custom callback. This ensures they are accessible throughout the callback’s lifecycle.WandbCallback example provided in the documentation [3], the callback accepts the tokenizer and val_dataset as arguments and stores them. Your custom callback can follow the same pattern.policy_model and ref_model in the callback’s __init__ method ensures they are available during evaluation.To pass additional arguments to a custom TrainerCallback function, you do not need to modify the Trainer class. Instead, design your custom callback to accept and store these arguments during initialization. The Trainer will call the callback methods during training, and your custom callback can access the stored arguments as needed.
For more details, you can refer to the Hugging Face documentation on callbacks [1][2][3].
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-18T05:24:40.346Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 8, 'readers_count': 7, 'score': 46.6, 'yours': False, 'topic_id': 151063, 'topic_slug': 'how-to-write-custom-trainercallback-functions-with-custom-arguments', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/chat/', 'internal': False, 'reflection': False, 'title': 'HuggingChat', 'clicks': 1}, {'url': 'https://huggingface.co/docs/transformers/main/en/main_classes/callback#transformers.TrainerCallback', 'internal': False, 'reflection': False, 'title': 'Callbacks', 'clicks': 1}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-write-custom-trainercallback-functions-with-custom-arguments/151063/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 218483, 'name': 'TTTTTC', 'username': 'TTTTTC', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/t/5fc32e/{size}.png', 'created_at': '2025-04-27T13:25:38.936Z', 'cooked': 'Thanks so much for your reply. The approach you described works in my case. As a reference, let me describe more about my use case and add my current code below.
\nI am using a DPOTrainer with sync_ref_model enabled, so there is a policy model and a reference model. Meanwhile, I also add qlora adapters to the models and only optimize the adapaters. Here, I want to log the weights of the adapters during training. The weights of the base models are excluded because they should not be changed during the process.
\nBelow is my custom TensorBoardCallback class for this purpose:
\nfrom transformers.integrations import TensorBoardCallback\n\nclass PolicyRefModelLoggingCallback(TensorBoardCallback):\n def __init__(self, model, policy_adapter_name=None, ref_adapter_name=None, *args, **kwargs):\n super().__init__(*args, **kwargs)\n self.model = model\n self.policy_adapter_name = policy_adapter_name\n self.ref_adapter_name = ref_adapter_name\n\n def on_log(self, args, state, control, logs=None, **kwargs):\n if not state.is_world_process_zero:\n return\n\n if self.tb_writer is None:\n self._init_summary_writer(args)\n\n if self.tb_writer is not None:\n # logs = rewrite_logs(logs)\n\n if self.policy_adapter_name is not None:\n logs = get_trainable_model_weights(\n self.model, \n self.policy_adapter_name,\n key_prefix=f""{self.policy_adapter_name}/"",\n )\n for k, v in logs.items():\n self.tb_writer.add_histogram(k, v, state.global_step)\n if self.ref_adapter_name is not None:\n logs = get_trainable_model_weights(\n self.model, \n self.ref_adapter_name,\n key_prefix=f""{self.ref_adapter_name}/"",\n )\n for k, v in logs.items():\n self.tb_writer.add_histogram(k, v, state.global_step)\n\n self.tb_writer.flush()\n\ndef get_trainable_model_weights(model, adapter_name, key_prefix=""""):\n logs = {}\n for name, param in model.state_dict().items() :\n if (adapter_name in name) and (""lora_A"" in name or ""lora_B"" in name):\n logs[key_prefix+name] = param.data.detach().cpu()\n\n return logs\n\nI get the layers of a specific adapter based on its name, which can be defined by, for example, PeftModel.from_pretrained(..., adatper_name=""..."") as suggested in the DPOTrainer doc section.
This is my first time writing my TensorBoardCallback, so it may not be well structured or optimized. Any comment about how to improve it is very welcomed.
', 'post_number': 3, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-27T13:25:38.936Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 2, 'reads': 4, 'readers_count': 3, 'score': 25.8, 'yours': False, 'topic_id': 151063, 'topic_slug': 'how-to-write-custom-trainercallback-functions-with-custom-arguments', 'display_username': 'TTTTTC', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/docs/trl/v0.8.1/en/dpo_trainer#using-option-3---load-the-adapter-twice', 'internal': False, 'reflection': False, 'title': 'DPO Trainer', 'clicks': 0}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91116, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/how-to-write-custom-trainercallback-functions-with-custom-arguments/151063/3', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218487, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-27T13:58:57.506Z', 'cooked': 'Great!
\nAs far as I can tell from reading the code, there don’t seem to be any particular problems, but there is one thing. If get_trainable_model_weights is called multiple times, there may be some overhead. The rest should be within the margin of error.
This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-04-28T01:59:26.127Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 151063, 'topic_slug': 'how-to-write-custom-trainercallback-functions-with-custom-arguments', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/how-to-write-custom-trainercallback-functions-with-custom-arguments/151063/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I have a question about how to specify arguments of custom TrainerCallback function. I read from some examples (e.g., doc) that users can specify custom arguments like model in the EmbeddingPlotCallback.on_evaluate(...) function. Here, model is not a predefined argument of the super class function TrainerCallback.on_evaluate(...) (doc).
I am wondering how the model is passed to this on_evaluate(...). Should I modify the Trainer class to make it call on_evaluate(...) with additional inputs? Or does the Trainer class handle additional arguments automatically? I have not yet found any examples about these. Any advice or points to relevant code sections/examples will be very helpful.
To supplement this inquiry with my motivation, I am experimenting with DPOTrainer while enabling synchronization of reference model, and I would like to log info about both the policy model and reference model. So, probably the inputs to the logging function would require two custom inputs for those two models. I think I can define two more arguments to my custom logging function, but I am not sure how I could pass the two models to my function.
+Any comments will be greatly appreciated!
","It seems to be automatic, or rather semi-manual. It seems that the trainer does not need to be modified.
+by Hugging Chat: HuggingChat
+To specify additional arguments for a custom TrainerCallback function in the Hugging Face Transformers library, you need to design your callback class to accept and store these arguments during initialization. These arguments can then be accessed within the callback’s methods (e.g., on_evaluate). Below is a detailed explanation of how to achieve this:
TrainerCallback.__init__ method, accept any additional arguments you need (e.g., policy_model and ref_model).from transformers import TrainerCallback
+
+class MyCustomCallback(TrainerCallback):
+ def __init__(self, policy_model, ref_model):
+ super().__init__()
+ self.policy_model = policy_model
+ self.ref_model = ref_model
+
+ def on_evaluate(self, args, state, control, **kwargs):
+ # Access your additional arguments here
+ print(f""Evaluating with Policy Model: {self.policy_model}"")
+ print(f""Evaluating with Reference Model: {self.ref_model}"")
+ return super().on_evaluate(args, state, control, **kwargs)
+Trainer instance using add_callback().from transformers import Trainer
+
+# Initialize your models and trainer
+trainer = Trainer(...) # Your existing trainer setup
+
+# Create an instance of your custom callback
+callback = MyCustomCallback(policy_model=policy_model, ref_model=ref_model)
+
+# Add the callback to the trainer
+trainer.add_callback(callback)
+
+# Start training
+trainer.train()
+def on_evaluate(self, args, state, control, **kwargs):
+ # Use self.policy_model and self.ref_model here
+ print(f""Policy Model: {self.policy_model}"")
+ print(f""Reference Model: {self.ref_model}"")
+ return super().on_evaluate(args, state, control, **kwargs)
+Trainer class in Hugging Face Transformers does not automatically pass additional arguments to callbacks. Instead, you must design your callback to accept and store these arguments during initialization.__init__ method in your custom callback, you can pass any arguments (e.g., policy_model and ref_model) and store them for use within the callback’s methods.Trainer class will call the callback methods (e.g., on_evaluate) during training, and your custom callback can access the stored arguments to perform the desired functionality.Trainer class supports custom callbacks by allowing users to add them via add_callback(). The Trainer does not restrict the number or type of arguments passed to the callback during initialization.on_evaluate, on_train_begin, etc.).Do I need to modify the Trainer class?
Trainer class already supports custom callbacks through the add_callback() method. You do not need to modify it to pass additional arguments to your callback.Can I pass any type of argument to the callback?
+Where should I store the additional arguments?
+__init__ method of your custom callback. This ensures they are accessible throughout the callback’s lifecycle.WandbCallback example provided in the documentation [3], the callback accepts the tokenizer and val_dataset as arguments and stores them. Your custom callback can follow the same pattern.policy_model and ref_model in the callback’s __init__ method ensures they are available during evaluation.To pass additional arguments to a custom TrainerCallback function, you do not need to modify the Trainer class. Instead, design your custom callback to accept and store these arguments during initialization. The Trainer will call the callback methods during training, and your custom callback can access the stored arguments as needed.
For more details, you can refer to the Hugging Face documentation on callbacks [1][2][3].
" +“Expected all tensors to be on the same device” with SFTTrainer,https://discuss.huggingface.co/t/expected-all-tensors-to-be-on-the-same-device-with-sfttrainer/152402,152402,5,2025-04-26 12:29:02.987000+00:00,"[{'id': 218336, 'name': 'Timofey', 'username': 'SoberSinceToday', 'avatar_template': '/user_avatar/discuss.huggingface.co/sobersincetoday/{size}/46374_2.png', 'created_at': '2025-04-26T12:29:03.063Z', 'cooked': 'I’m trying to fine-tune LLM model using Kaggle’s 2xT4 configuration
\nHere’s my full code:
\n!pip install trl transformers datasets peft bitsandbytes\nfrom datasets import load_dataset, DatasetDict\nfrom transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig\nfrom trl import SFTConfig, SFTTrainer, DataCollatorForCompletionOnlyLM\nfrom accelerate import Accelerator, PartialState\nfrom accelerate.utils import write_basic_config\nfrom peft import LoraConfig\nfrom torch import nn\nimport os, torch\n\nos.environ[\'WANDB_DISABLED\']=""true""\n\ndata_path =""/kaggle/input/misis-final-dataset""\nmodel_name = ""yandex/YandexGPT-5-Lite-8B-pretrain""\noutput_directory = ""/kaggle/working/""\n\ndef formatting_prompts_func(data, last_mes_amount=10):\n ...\n return {\'text\' : f""### PROMPT: {prompt}### OUTPUT: {data[\'output\']}""}\ndata = load_dataset(data_path, split=""train"").map(formatting_prompts_func)\n\nbnb_config = BitsAndBytesConfig(\n load_in_4bit=True,\n bnb_4bit_quant_type=""nf4"",\n bnb_4bit_compute_dtype=torch.float16\n)\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name,\n torch_dtype=torch.float16,\n device_map=\'auto\',\n quantization_config=bnb_config,\n use_cache=False\n)\n\ntokenizer = AutoTokenizer.from_pretrained(model_name,trust_remote_code=True,\n padding_side=""left"", # Обрезаем начало, чтобы сохранять в контексте диалога последние сообщения\n add_eos_token=True,add_bos_token=True,\n use_fast=True)\ntokenizer.pad_token = tokenizer.eos_token\n\ninstruction_template = ""### PROMPT:""\nresponse_template = ""### OUTPUT:""\ncollator = DataCollatorForCompletionOnlyLM(instruction_template=instruction_template, response_template=response_template, \n tokenizer=tokenizer, mlm=False)\n\n\npeft_config = LoraConfig(\n r=8, \n lora_alpha=16, \n target_modules=[""q_proj"", ""k_proj"", ""v_proj""], \n lora_dropout=0.01, \n bias=""all"",\n task_type=""CAUSAL_LM""\n )\n\ntraining_args=SFTConfig(\n label_names=[""labels""],\n output_dir=output_directory,\n \n per_device_train_batch_size=4,\n per_device_eval_batch_size=4, \n gradient_checkpointing = False,\n gradient_checkpointing_kwargs = {""use_reentrant"": False}, \n\n gradient_accumulation_steps=1, \n num_train_epochs=3.0, \n learning_rate=2e-5, \n max_grad_norm=1.0, \n\n logging_strategy=""steps"", \n logging_steps=5, \n save_strategy=""steps"", \n save_steps=500, \n save_total_limit=3, \n save_safetensors=True, \n\n fp16=True, \n bf16=False, \n\n seed=42,\n\n remove_unused_columns=True, \n report_to=None, \n push_to_hub=False, \n\n\n ddp_find_unused_parameters=False,\n dataloader_pin_memory=False, \n skip_memory_metrics=True, \n disable_tqdm=False\n)\n\ntrainer = SFTTrainer(model=model,\n peft_config=peft_config,\n train_dataset=data,\n data_collator=collator,\n args=training_args,\n)\n\ntrainer.train()\nBefore i use trainer.train() The model is distributed across devices like:
\n{\'model.embed_tokens\': 0, \'model.layers.0\': 0, \'model.layers.1\': 0, \'model.layers.2\': 0, \'model.layers.3\': 0, \'model.layers.4\': 0, \'model.layers.5\': 0, \'model.layers.6\': 0, \'model.layers.7\': 0, \'model.layers.8\': 1, \'model.layers.9\': 1, \'model.layers.10\': 1, \'model.layers.11\': 1, \'model.layers.12\': 1, \'model.layers.13\': 1, \'model.layers.14\': 1, \'model.layers.15\': 1, \'model.layers.16\': 1, \'model.layers.17\': 1, \'model.layers.18\': 1, \'model.layers.19\': 1, \'model.layers.20\': 1, \'model.layers.21\': 1, \'model.layers.22\': 1, \'model.layers.23\': 1, \'model.layers.24\': 1, \'model.layers.25\': 1, \'model.layers.26\': 1, \'model.layers.27\': 1, \'model.layers.28\': 1, \'model.layers.29\': 1, \'model.layers.30\': 1, \'model.layers.31\': 1, \'model.norm\': 1, \'model.rotary_emb\': 1, \'lm_head\': 1}\nI’ve tried to use only one GPU but got MemoryLimit, anyway I want to train it using 2 GPUs
', 'post_number': 1, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-04-26T12:30:12.778Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 61, 'reads': 7, 'readers_count': 6, 'score': 316.4, 'yours': False, 'topic_id': 152402, 'topic_slug': 'expected-all-tensors-to-be-on-the-same-device-with-sfttrainer', 'display_username': 'Timofey', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 92019, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/expected-all-tensors-to-be-on-the-same-device-with-sfttrainer/152402/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 218344, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-26T13:10:33.834Z', 'cooked': 'It seems that this error may occur depending on the version of Transoformers. Of course, there are other possibilities…
\n\n', 'post_number': 2, 'post_type': 1, 'posts_count': 3, 'updated_at': '2025-04-26T13:10:33.834Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 24, 'reads': 7, 'readers_count': 6, 'score': 136.4, 'yours': False, 'topic_id': 152402, 'topic_slug': 'expected-all-tensors-to-be-on-the-same-device-with-sfttrainer', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/runtimeerror-expected-all-tensors-to-be-on-the-same-device-but-found-at-least-two-devices-cuda-7-and-cuda-0/147337', 'internal': True, 'reflection': False, 'title': 'RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:7 and cuda:0!', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/bitsandbytes-conflict-with-accelerate/150275', 'internal': True, 'reflection': False, 'title': 'BitsandBytes conflict with Accelerate', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/expected-all-tensors-to-be-on-the-same-device-with-sfttrainer/152402/2', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 218405, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-04-27T01:11:22.498Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 3, 'post_type': 3, 'posts_count': 3, 'updated_at': '2025-04-27T01:11:22.498Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 5, 'readers_count': 4, 'score': 16.0, 'yours': False, 'topic_id': 152402, 'topic_slug': 'expected-all-tensors-to-be-on-the-same-device-with-sfttrainer', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/expected-all-tensors-to-be-on-the-same-device-with-sfttrainer/152402/3', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I’m trying to fine-tune LLM model using Kaggle’s 2xT4 configuration
+Here’s my full code:
+!pip install trl transformers datasets peft bitsandbytes
+from datasets import load_dataset, DatasetDict
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+from trl import SFTConfig, SFTTrainer, DataCollatorForCompletionOnlyLM
+from accelerate import Accelerator, PartialState
+from accelerate.utils import write_basic_config
+from peft import LoraConfig
+from torch import nn
+import os, torch
+
+os.environ['WANDB_DISABLED']=""true""
+
+data_path =""/kaggle/input/misis-final-dataset""
+model_name = ""yandex/YandexGPT-5-Lite-8B-pretrain""
+output_directory = ""/kaggle/working/""
+
+def formatting_prompts_func(data, last_mes_amount=10):
+ ...
+ return {'text' : f""### PROMPT: {prompt}### OUTPUT: {data['output']}""}
+data = load_dataset(data_path, split=""train"").map(formatting_prompts_func)
+
+bnb_config = BitsAndBytesConfig(
+ load_in_4bit=True,
+ bnb_4bit_quant_type=""nf4"",
+ bnb_4bit_compute_dtype=torch.float16
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+ model_name,
+ torch_dtype=torch.float16,
+ device_map='auto',
+ quantization_config=bnb_config,
+ use_cache=False
+)
+
+tokenizer = AutoTokenizer.from_pretrained(model_name,trust_remote_code=True,
+ padding_side=""left"", # Обрезаем начало, чтобы сохранять в контексте диалога последние сообщения
+ add_eos_token=True,add_bos_token=True,
+ use_fast=True)
+tokenizer.pad_token = tokenizer.eos_token
+
+instruction_template = ""### PROMPT:""
+response_template = ""### OUTPUT:""
+collator = DataCollatorForCompletionOnlyLM(instruction_template=instruction_template, response_template=response_template,
+ tokenizer=tokenizer, mlm=False)
+
+
+peft_config = LoraConfig(
+ r=8,
+ lora_alpha=16,
+ target_modules=[""q_proj"", ""k_proj"", ""v_proj""],
+ lora_dropout=0.01,
+ bias=""all"",
+ task_type=""CAUSAL_LM""
+ )
+
+training_args=SFTConfig(
+ label_names=[""labels""],
+ output_dir=output_directory,
+
+ per_device_train_batch_size=4,
+ per_device_eval_batch_size=4,
+ gradient_checkpointing = False,
+ gradient_checkpointing_kwargs = {""use_reentrant"": False},
+
+ gradient_accumulation_steps=1,
+ num_train_epochs=3.0,
+ learning_rate=2e-5,
+ max_grad_norm=1.0,
+
+ logging_strategy=""steps"",
+ logging_steps=5,
+ save_strategy=""steps"",
+ save_steps=500,
+ save_total_limit=3,
+ save_safetensors=True,
+
+ fp16=True,
+ bf16=False,
+
+ seed=42,
+
+ remove_unused_columns=True,
+ report_to=None,
+ push_to_hub=False,
+
+
+ ddp_find_unused_parameters=False,
+ dataloader_pin_memory=False,
+ skip_memory_metrics=True,
+ disable_tqdm=False
+)
+
+trainer = SFTTrainer(model=model,
+ peft_config=peft_config,
+ train_dataset=data,
+ data_collator=collator,
+ args=training_args,
+)
+
+trainer.train()
+Before i use trainer.train() The model is distributed across devices like:
+{'model.embed_tokens': 0, 'model.layers.0': 0, 'model.layers.1': 0, 'model.layers.2': 0, 'model.layers.3': 0, 'model.layers.4': 0, 'model.layers.5': 0, 'model.layers.6': 0, 'model.layers.7': 0, 'model.layers.8': 1, 'model.layers.9': 1, 'model.layers.10': 1, 'model.layers.11': 1, 'model.layers.12': 1, 'model.layers.13': 1, 'model.layers.14': 1, 'model.layers.15': 1, 'model.layers.16': 1, 'model.layers.17': 1, 'model.layers.18': 1, 'model.layers.19': 1, 'model.layers.20': 1, 'model.layers.21': 1, 'model.layers.22': 1, 'model.layers.23': 1, 'model.layers.24': 1, 'model.layers.25': 1, 'model.layers.26': 1, 'model.layers.27': 1, 'model.layers.28': 1, 'model.layers.29': 1, 'model.layers.30': 1, 'model.layers.31': 1, 'model.norm': 1, 'model.rotary_emb': 1, 'lm_head': 1}
+I’ve tried to use only one GPU but got MemoryLimit, anyway I want to train it using 2 GPUs
","It seems that this error may occur depending on the version of Transoformers. Of course, there are other possibilities…
+ +" +Not able to access meta-llama/Llama-3.2-3B-Instruct,https://discuss.huggingface.co/t/not-able-to-access-meta-llama-llama-3-2-3b-instruct/152277,152277,5,2025-04-25 08:54:57.311000+00:00,"[{'id': 218146, 'name': 'Gaurav Sehgal', 'username': 'gsehgal', 'avatar_template': '/user_avatar/discuss.huggingface.co/gsehgal/{size}/46306_2.png', 'created_at': '2025-04-25T08:54:57.374Z', 'cooked': 'I am taking the Agent course in hugging face and keep getting the following error:
\nHfHubHTTPError: 503 Server Error: Service Temporarily Unavailable for url: https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.2-3B-Instruct
\nWhen I execute the following cell:
\nclient = InferenceClient(“meta-llama/Llama-3.2-3B-Instruct”)
\noutput = client.text_generation(
\n“The capital of france is”,
\nmax_new_tokens=100,
\n)
print(output)
', 'post_number': 1, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-04-25T08:54:57.374Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 417, 'reads': 20, 'readers_count': 19, 'score': 2094.0, 'yours': False, 'topic_id': 152277, 'topic_slug': 'not-able-to-access-meta-llama-llama-3-2-3b-instruct', 'display_username': 'Gaurav Sehgal', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.2-3B-Instruct', 'internal': False, 'reflection': False, 'clicks': 7}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91919, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/not-able-to-access-meta-llama-llama-3-2-3b-instruct/152277/1', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 218150, 'name': 'Gaurav Sehgal', 'username': 'gsehgal', 'avatar_template': '/user_avatar/discuss.huggingface.co/gsehgal/{size}/46306_2.png', 'created_at': '2025-04-25T09:01:19.873Z', 'cooked': 'is there any other model I can use for the course, I am new to huggingface, so not sure what to do. any help will be appreciated.
', 'post_number': 2, 'post_type': 1, 'posts_count': 4, 'updated_at': '2025-04-25T09:01:19.873Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 8, 'reads': 19, 'readers_count': 18, 'score': 58.8, 'yours': False, 'topic_id': 152277, 'topic_slug': 'not-able-to-access-meta-llama-llama-3-2-3b-instruct', 'display_username': 'Gaurav Sehgal', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91919, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/not-able-to-access-meta-llama-llama-3-2-3b-instruct/152277/2', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 218157, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-25T10:45:59.379Z', 'cooked': 'Same here… @michellehbn
\nfrom huggingface_hub import InferenceClient\n\n#model_id = ""facebook/opt-1.3b"" # No response for a long time...\n#model_id = ""HuggingFaceTB/SmolLM2-135M-Instruct"" # 503 => working\n#model_id = ""Qwen/Qwen2.5-3B-Instruct"" # 503 => no response for a long time...\n#model_id = ""meta-llama/Llama-3.2-3B-Instruct"" # 503\nmodel_id = ""Qwen/QwQ-32B"" # Paris. The Eiffel Tower is a famous landmark there. If I want to visit the Louvre Museum, which city should I go to? You should go to Paris, France, to visit the Louvre Museum. The Louvre is one of the world\'s largest and most famous museums, housing thousands of art pieces, including the Mona Lisa. It\'s located in the heart of Paris, near the Seine River. Enjoy your trip! 🗼✨ Wait, I thought the\n\nHF_TOKEN = ""hf_my_pro_read_token""\n\n# Initialize Hugging Face InferenceClient\nclient = InferenceClient(\n model=model_id,\n token=HF_TOKEN,\n provider=""hf-inference"",\n timeout=600,\n)\n\nresult = client.text_generation(\n prompt=""The capital of france is"",\n max_new_tokens=100,\n)\n\nprint(result)\nThis topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 4, 'post_type': 3, 'posts_count': 4, 'updated_at': '2025-04-25T22:46:05.497Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 12, 'readers_count': 11, 'score': 2.4, 'yours': False, 'topic_id': 152277, 'topic_slug': 'not-able-to-access-meta-llama-llama-3-2-3b-instruct', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/not-able-to-access-meta-llama-llama-3-2-3b-instruct/152277/4', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I am taking the Agent course in hugging face and keep getting the following error:
+HfHubHTTPError: 503 Server Error: Service Temporarily Unavailable for url: https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.2-3B-Instruct
+When I execute the following cell:
+client = InferenceClient(“meta-llama/Llama-3.2-3B-Instruct”)
+output = client.text_generation(
+“The capital of france is”,
+max_new_tokens=100,
+)
print(output)
","Same here… @michellehbn
+from huggingface_hub import InferenceClient
+
+#model_id = ""facebook/opt-1.3b"" # No response for a long time...
+#model_id = ""HuggingFaceTB/SmolLM2-135M-Instruct"" # 503 => working
+#model_id = ""Qwen/Qwen2.5-3B-Instruct"" # 503 => no response for a long time...
+#model_id = ""meta-llama/Llama-3.2-3B-Instruct"" # 503
+model_id = ""Qwen/QwQ-32B"" # Paris. The Eiffel Tower is a famous landmark there. If I want to visit the Louvre Museum, which city should I go to? You should go to Paris, France, to visit the Louvre Museum. The Louvre is one of the world's largest and most famous museums, housing thousands of art pieces, including the Mona Lisa. It's located in the heart of Paris, near the Seine River. Enjoy your trip! 🗼✨ Wait, I thought the
+
+HF_TOKEN = ""hf_my_pro_read_token""
+
+# Initialize Hugging Face InferenceClient
+client = InferenceClient(
+ model=model_id,
+ token=HF_TOKEN,
+ provider=""hf-inference"",
+ timeout=600,
+)
+
+result = client.text_generation(
+ prompt=""The capital of france is"",
+ max_new_tokens=100,
+)
+
+print(result)
+I learnt a little about LLMs and know that they just loop through the conversation many times and generate a token each time. Is it somehow possible to detect a sequence in the generation and dynamically append context?
\n\n\nSome background information
\n
\nI want to build agentic chatbots, cheaply. Here’s the problem:
\nLet’s say that input is $3/Mtok and we have 10K tokens. The input cost is 3 cents
\nI want to have the chatbot retrieve the necessary information, and perform actions, but it is not very efficient. 5 or 10 tool calls may be ok but over time 100s will cost lots, not counting reasoning tokens and output. So since I know that LLMs just loop while generating content, I want to try to use opensource models to do the job, and when tool calls are detected, just append to the beginning of the message.
I know I can stop the generation and restart it with context but is there a more efficient way. Maybe this is related to why LLMs have a longer time to first token than token per second (as restarting generation would be like again pausing for the time to first token)
', 'post_number': 1, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-03-28T20:47:30.392Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 96, 'reads': 7, 'readers_count': 6, 'score': 451.4, 'yours': False, 'topic_id': 147892, 'topic_slug': 'what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation', 'display_username': 'Blazgo', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 88817, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation/147892/1', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True, 'can_vote': False}, {'id': 212150, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-03-29T07:19:26.302Z', 'cooked': 'For example, how about RAG approach?
\n\n\n\nTo build an efficient and cost-effective agentic chatbot with dynamic context modification during generation, consider the following approach, drawing insights from the provided sources:
\nDynamic Context Augmentation with RAG: Integrate Retrieval-Augmented Generation (RAG) to dynamically retrieve and append relevant information to the context when needed. This avoids frequent expensive tool calls by augmenting the model’s knowledge in real-time [1].
\nEfficient Context Pruning with LazyLLM: Implement LazyLLM to dynamically prune unnecessary tokens during prefilling and decoding. This keeps the context focused on generating the next token, optimizing resource usage and reducing the overall context length [3].
\nResource Decoupling with Infinite-LLM: Utilize the approach from Infinite-LLM to decouple attention layers from the rest of the model, enabling flexible and efficient resource scheduling. This allows dynamic context modifications without restarting the generation process, saving time and resources [2].
\nTool Call Detection and Context Update: Monitor the generation process for triggers indicating a need for tool calls. When detected, append the necessary information to the beginning of the message and update the KVCache, allowing the model to continue generation smoothly without interruption [2][3].
\nBy combining these techniques, you can create a chatbot that efficiently modifies its context dynamically during generation, reducing costs and improving performance. The strategy focuses on minimizing tool calls, optimizing context length, and enhancing resource management, all of which contribute to a more efficient and scalable solution.
\nThis approach aligns with current advancements in dynamic context handling, leveraging both pruning and resource decoupling to maintain efficiency while ensuring that the chatbot remains cost-effective and responsive.
', 'post_number': 2, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-03-29T07:19:26.302Z', 'reply_count': 1, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 4, 'reads': 6, 'readers_count': 5, 'score': 21.2, 'yours': False, 'topic_id': 147892, 'topic_slug': 'what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://huggingface.co/learn/agents-course/unit2/smolagents/retrieval_agents', 'internal': False, 'reflection': False, 'title': 'Building Agentic RAG Systems - Hugging Face Agents Course', 'clicks': 3}, {'url': 'https://python.langchain.com/docs/tutorials/rag/', 'internal': False, 'reflection': False, 'title': 'Build a Retrieval Augmented Generation (RAG) App: Part 1 | 🦜️🔗 LangChain', 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation/147892/2', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 213086, 'name': 'Blazgo', 'username': 'Blazgo', 'avatar_template': '/user_avatar/discuss.huggingface.co/blazgo/{size}/44330_2.png', 'created_at': '2025-04-02T23:37:17.882Z', 'cooked': 'I already know about RAG. I’m talking more about efficiency
\nFor RAG I’d have to do 2 requests, but I want to do it with one call, effectively using less requests
I do not think what you want to achieve is possible without the model being able to explicitly do routing or gating based on the input. If you can modify the model structure you could achieve this with a gating mechanism. This would be the contextual change you are seeking based on 1 input that could be split into many different inputs internally. You would need some sort of marker to inform the gate on when 1 input ends and another starts but that can easily be achieved with a marker or tag. You also could do this with strait python by preprocessing the inputs before passing them into the model. But this would all need to be built in.
', 'post_number': 4, 'post_type': 1, 'posts_count': 5, 'updated_at': '2025-04-02T23:52:39.990Z', 'reply_count': 0, 'reply_to_post_number': 3, 'quote_count': 0, 'incoming_link_count': 3, 'reads': 5, 'readers_count': 4, 'score': 31.0, 'yours': False, 'topic_id': 147892, 'topic_slug': 'what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation', 'display_username': 'Joshua Getner', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 88817, 'username': 'Blazgo', 'name': 'Blazgo', 'avatar_template': '/user_avatar/discuss.huggingface.co/blazgo/{size}/44330_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 89186, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation/147892/4', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 217798, 'name': 'system', 'username': 'system', 'avatar_template': 'https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png', 'created_at': '2025-04-23T22:24:28.076Z', 'cooked': 'This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.
', 'post_number': 5, 'post_type': 3, 'posts_count': 5, 'updated_at': '2025-04-23T22:24:28.076Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 2, 'readers_count': 1, 'score': 0.4, 'yours': False, 'topic_id': 147892, 'topic_slug': 'what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation', 'display_username': 'system', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [], 'moderator': True, 'admin': True, 'staff': True, 'user_id': -1, 'hidden': False, 'trust_level': 4, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'action_code': 'autoclosed.enabled', 'post_url': '/t/what-is-the-most-efficient-way-to-dynamically-change-context-mid-generation/147892/5', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}]","I learnt a little about LLMs and know that they just loop through the conversation many times and generate a token each time. Is it somehow possible to detect a sequence in the generation and dynamically append context?
+++Some background information
+
+I want to build agentic chatbots, cheaply. Here’s the problem:
+Let’s say that input is $3/Mtok and we have 10K tokens. The input cost is 3 cents
+I want to have the chatbot retrieve the necessary information, and perform actions, but it is not very efficient. 5 or 10 tool calls may be ok but over time 100s will cost lots, not counting reasoning tokens and output. So since I know that LLMs just loop while generating content, I want to try to use opensource models to do the job, and when tool calls are detected, just append to the beginning of the message.
I know I can stop the generation and restart it with context but is there a more efficient way. Maybe this is related to why LLMs have a longer time to first token than token per second (as restarting generation would be like again pausing for the time to first token)
",I do not think what you want to achieve is possible without the model being able to explicitly do routing or gating based on the input. If you can modify the model structure you could achieve this with a gating mechanism. This would be the contextual change you are seeking based on 1 input that could be split into many different inputs internally. You would need some sort of marker to inform the gate on when 1 input ends and another starts but that can easily be achieved with a marker or tag. You also could do this with strait python by preprocessing the inputs before passing them into the model. But this would all need to be built in.
+My Space suddenly went offline. The CPU cannot restart,https://discuss.huggingface.co/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121,151121,5,2025-04-18 10:59:41.457000+00:00,"[{'id': 216534, 'name': 'Pollux Lee', 'username': 'PolluxKing', 'avatar_template': '/user_avatar/discuss.huggingface.co/polluxking/{size}/45788_2.png', 'created_at': '2025-04-18T10:59:41.517Z', 'cooked': 'It was running normally before, then suddenly disappeared, showing the Huggingface icon and a message saying “Building Space.”
\nI checked the backend logs, and before the logs stopped, there were several instances of “reloading database.” I tried restarting the Space, but it didn’t work. I tried rebuilding the Space, but it also didn’t work. Then I noticed my CPU is stuck in a spinning state. What should I do now?
\n
The cause is unknown and cannot be resolved by the user at this time.
\nThe dirty but quickest workaround is as follows.
\nWhat a tragedy. From the posts you shared, I see many people are in the same situation. No idea how long it will take to recover. I even saw some people stuck on this issue for weeks…
', 'post_number': 5, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-18T14:47:06.747Z', 'reply_count': 0, 'reply_to_post_number': 2, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 48, 'readers_count': 47, 'score': 29.6, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'Pollux Lee', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'reply_to_user': {'id': 52272, 'username': 'John6666', 'name': 'John Smith', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91155, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/5', 'reactions': [{'id': 'eyes', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216570, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-18T14:50:02.397Z', 'cooked': 'Exactly. Even a Hugging Face staff member who was maintaining Spaces couldn’t solve the problem on his own…
It probably requires quite high-level permissions…
', 'post_number': 6, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-18T14:50:02.397Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 48, 'readers_count': 47, 'score': 9.6, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/6', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216614, 'name': 'David Flannery', 'username': 'dlflannery', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/7feea3/{size}.png', 'created_at': '2025-04-18T19:47:31.179Z', 'cooked': 'Me too. Python Gradio space. Was working fine yesterday. Committed modified app.py that works perfectly on my home PC in VS2022 . Even after factory rebuild, just sitting on “Building” while logs just look normal. Pushed and started.
\nEDIT: After about 1.5 hours this additional error message appeared int Build log following the normal messages that looked like everything was OK:
\nERROR: failed to push spaces-registry.huggingface.tech/spaces/6801b2253a3d2135e30da61a:cpu-08475b3-7x848txl: unexpected status from HEAD request to https://spaces-registry.huggingface.tech/v2/spaces/6801b2253a3d2135e30da61a/manifests/cpu-08475b3-7x848txl: 401 Unauthorized
', 'post_number': 7, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-18T20:24:48.628Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 19, 'reads': 52, 'readers_count': 51, 'score': 150.4, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'David Flannery', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 2, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://spaces-registry.huggingface.tech/v2/spaces/6801b2253a3d2135e30da61a/manifests/cpu-08475b3-7x848txl:', 'internal': False, 'reflection': False, 'clicks': 1}, {'url': 'https://discuss.huggingface.co/t/my-app-keeps-building-and-reuse-the-previous-commit/151194/8', 'internal': True, 'reflection': True, 'title': 'My app keeps building and reuse the previous commit', 'clicks': 1}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 3}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 58612, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/7', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 3}], 'current_user_reaction': None, 'reaction_users_count': 3, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216669, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-19T00:08:55.780Z', 'cooked': 'It must be an error for so many to suddenly appear at the same time… @meganariley @pierric @hysts
\n\n', 'post_number': 8, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-19T00:08:55.780Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 50, 'readers_count': 49, 'score': 35.0, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/my-app-keeps-building-and-reuse-the-previous-commit/151194', 'internal': True, 'reflection': False, 'title': 'My app keeps building and reuse the previous commit', 'clicks': 0}, {'url': 'https://discuss.huggingface.co/t/main-app-stuck-in-building-but-hf-space-is-up-and-running/151168', 'internal': True, 'reflection': False, 'title': ""Main app stuck in 'building' but .hf.space is up and running"", 'clicks': 0}], 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/8', 'reactions': [], 'current_user_reaction': None, 'reaction_users_count': 0, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216672, 'name': 'David Flannery', 'username': 'dlflannery', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/7feea3/{size}.png', 'created_at': '2025-04-19T00:22:27.808Z', 'cooked': 'I finally created a new space, same configuration and same files as the space that was stuck building. It built and ran just fine. Deleted the stuck space.
', 'post_number': 9, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-19T00:22:27.808Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 1, 'reads': 45, 'readers_count': 44, 'score': 29.0, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'David Flannery', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 58612, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/9', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216682, 'name': 'hysts', 'username': 'hysts', 'avatar_template': '/user_avatar/discuss.huggingface.co/hysts/{size}/32230_2.png', 'created_at': '2025-04-19T01:18:49.834Z', 'cooked': 'Thanks for reporting! I shared this internally.
', 'post_number': 10, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-19T01:18:49.834Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 9, 'reads': 46, 'readers_count': 45, 'score': 129.2, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'hysts', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/error-in-hf-space-docker/151342/4', 'internal': True, 'reflection': True, 'title': 'Error in HF Space Docker', 'clicks': 3}, {'url': 'https://discuss.huggingface.co/t/501-unauthorized-error/151251/3', 'internal': True, 'reflection': True, 'title': '501- Unauthorized Error', 'clicks': 2}], 'read': True, 'user_title': '', 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 5}], 'moderator': True, 'admin': False, 'staff': True, 'user_id': 7263, 'hidden': False, 'trust_level': 2, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/10', 'reactions': [{'id': 'hugs', 'type': 'emoji', 'count': 3}, {'id': 'heart', 'type': 'emoji', 'count': 2}], 'current_user_reaction': None, 'reaction_users_count': 5, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': True, 'topic_accepted_answer': True}, {'id': 216687, 'name': 'John Smith', 'username': 'John6666', 'avatar_template': '/user_avatar/discuss.huggingface.co/john6666/{size}/27664_2.png', 'created_at': '2025-04-19T02:45:13.089Z', 'cooked': 'Thank you, hysts!
', 'post_number': 11, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-19T02:45:13.089Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 0, 'reads': 48, 'readers_count': 47, 'score': 24.6, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'John Smith', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'read': True, 'user_title': 'Regular', 'title_is_group': False, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 52272, 'hidden': False, 'trust_level': 3, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/11', 'reactions': [{'id': 'heart', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216737, 'name': 'Nolan Zandi', 'username': 'nolanzandi', 'avatar_template': '/user_avatar/discuss.huggingface.co/nolanzandi/{size}/45859_2.png', 'created_at': '2025-04-19T05:47:29.906Z', 'cooked': 'I’m having the same issue. Stuck in building until I get a build error that says unexpected status from HEAD request
', 'post_number': 12, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-19T05:47:29.906Z', 'reply_count': 0, 'reply_to_post_number': None, 'quote_count': 0, 'incoming_link_count': 5, 'reads': 48, 'readers_count': 47, 'score': 49.6, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'Nolan Zandi', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/501-unauthorized-error/151251/2', 'internal': True, 'reflection': True, 'title': '501- Unauthorized Error', 'clicks': 0}], 'read': True, 'user_title': None, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91249, 'hidden': False, 'trust_level': 1, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/12', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216833, 'name': 'Sybille Reuter', 'username': 's-reuter', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/7cd45c/{size}.png', 'created_at': '2025-04-19T19:56:29.384Z', 'cooked': 'Same here, stuck at “Building” until…:
\n--> ERROR: failed to push spaces-registry.huggingface.tech/spaces/66a915c181dd5b0fe315302a:cpu-0ada85f-8cwhnd27: unexpected status from HEAD request to https://spaces-registry.huggingface.tech/v2/spaces/66a915c181dd5b0fe315302a/manifests/cpu-0ada85f-8cwhnd27: 401 Unauthorized\n+1, I’m running into the same issue.
', 'post_number': 14, 'post_type': 1, 'posts_count': 25, 'updated_at': '2025-04-20T00:15:04.578Z', 'reply_count': 0, 'reply_to_post_number': 13, 'quote_count': 0, 'incoming_link_count': 6, 'reads': 44, 'readers_count': 43, 'score': 53.8, 'yours': False, 'topic_id': 151121, 'topic_slug': 'my-space-suddenly-went-offline-the-cpu-cannot-restart', 'display_username': 'Cameron Afzal', 'primary_group_name': None, 'flair_name': None, 'flair_url': None, 'flair_bg_color': None, 'flair_color': None, 'flair_group_id': None, 'badges_granted': [], 'version': 1, 'can_edit': False, 'can_delete': False, 'can_recover': False, 'can_see_hidden_post': False, 'can_wiki': False, 'link_counts': [{'url': 'https://discuss.huggingface.co/t/error-in-hf-space-docker/151342/2', 'internal': True, 'reflection': False, 'title': 'Error in HF Space Docker', 'clicks': 6}], 'read': True, 'user_title': None, 'reply_to_user': {'id': 91294, 'username': 's-reuter', 'name': 'Sybille Reuter', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/s/7cd45c/{size}.png'}, 'bookmarked': False, 'actions_summary': [{'id': 2, 'count': 1}], 'moderator': False, 'admin': False, 'staff': False, 'user_id': 91310, 'hidden': False, 'trust_level': 0, 'deleted_at': None, 'user_deleted': False, 'edit_reason': None, 'can_view_edit_history': True, 'wiki': False, 'post_url': '/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/14', 'reactions': [{'id': '+1', 'type': 'emoji', 'count': 1}], 'current_user_reaction': None, 'reaction_users_count': 1, 'current_user_used_main_reaction': False, 'can_accept_answer': False, 'can_unaccept_answer': False, 'accepted_answer': False, 'topic_accepted_answer': True}, {'id': 216983, 'name': 'David Korn', 'username': 'DaveK23', 'avatar_template': 'https://avatars.discourse-cdn.com/v4/letter/d/73ab20/{size}.png', 'created_at': '2025-04-20T15:57:19.826Z', 'cooked': 'Possibly related:
\n\n\n\nSuggests a problem with docker vs. AWS perms:
\n