Add Parquet files (#2)

- Rename loader script (618096d8057a6aba10c1119e018c277cedb67715)
- Upload parquet/all (39c588fc7824d08c980a01c48d2502b5a461550a)
- Upload parquet/en (f779a4245bb273d4e4a82306fd481259bcf35e14)
- Upload parquet/es (107fa1fd88788f8e99b3ac1d567e0c5d968b4405)
- Upload parquet/de (53e3e9440d26e82ab8c7459fd06cf53088187865)
- Upload parquet/fr (25c746f96e62b118228bb59b593e14114e8b7dcf)
- Add push_to_hub.py (27752fa76f63e95191296428424aa339676a3200)
- Update README.md (5b65248c8862079c7d19f3765abbb945aa6f5755)
- Version 1.1.0 (db8d62101bfc3d833dcfb2a5fbebb33cd3110bb5)

README.md

@@ -16,6 +16,119 @@ tags:
 - transcription
 pretty_name: 'JamALT: A Readability-Aware Lyrics Transcription Benchmark'
 paperswithcode_id: jam-alt
+dataset_info:
+- config_name: all
+  features:
+  - name: name
+    dtype: string
+  - name: text
+    dtype: string
+  - name: language
+    dtype: string
+  - name: license_type
+    dtype: string
+  - name: audio
+    dtype: audio
+  splits:
+  - name: test
+    num_bytes: 409411912.0
+    num_examples: 79
+  download_size: 409150043
+  dataset_size: 409411912.0
+- config_name: de
+  features:
+  - name: name
+    dtype: string
+  - name: text
+    dtype: string
+  - name: language
+    dtype: string
+  - name: license_type
+    dtype: string
+  - name: audio
+    dtype: audio
+  splits:
+  - name: test
+    num_bytes: 107962802.0
+    num_examples: 20
+  download_size: 107942102
+  dataset_size: 107962802.0
+- config_name: en
+  features:
+  - name: name
+    dtype: string
+  - name: text
+    dtype: string
+  - name: language
+    dtype: string
+  - name: license_type
+    dtype: string
+  - name: audio
+    dtype: audio
+  splits:
+  - name: test
+    num_bytes: 105135091.0
+    num_examples: 20
+  download_size: 105041371
+  dataset_size: 105135091.0
+- config_name: es
+  features:
+  - name: name
+    dtype: string
+  - name: text
+    dtype: string
+  - name: language
+    dtype: string
+  - name: license_type
+    dtype: string
+  - name: audio
+    dtype: audio
+  splits:
+  - name: test
+    num_bytes: 105024257.0
+    num_examples: 20
+  download_size: 104979012
+  dataset_size: 105024257.0
+- config_name: fr
+  features:
+  - name: name
+    dtype: string
+  - name: text
+    dtype: string
+  - name: language
+    dtype: string
+  - name: license_type
+    dtype: string
+  - name: audio
+    dtype: audio
+  splits:
+  - name: test
+    num_bytes: 91289764.0
+    num_examples: 19
+  download_size: 91218543
+  dataset_size: 91289764.0
+configs:
+- config_name: all
+  data_files:
+  - split: test
+    path: parquet/all/test-*
+  default: true
+- config_name: de
+  data_files:
+  - split: test
+    path: parquet/de/test-*
+- config_name: en
+  data_files:
+  - split: test
+    path: parquet/en/test-*
+- config_name: es
+  data_files:
+  - split: test
+    path: parquet/es/test-*
+- config_name: fr
+  data_files:
+  - split: test
+    path: parquet/fr/test-*
 ---
 
 # JamALT: A Readability-Aware Lyrics Transcription Benchmark
@@ -42,17 +155,18 @@ See the [project website](https://audioshake.github.io/jam-alt/) for details.
 
 ```python
 from datasets import load_dataset
-dataset = load_dataset("audioshake/jam-alt")["test"]
+dataset = load_dataset("audioshake/jam-alt", split="test")
 ```
 
 A subset is defined for each language (`en`, `fr`, `de`, `es`);
 for example, use `load_dataset("audioshake/jam-alt", "es")` to load only the Spanish songs.
 
-By default, the dataset comes with audio. To skip loading the audio, use `with_audio=False`.
 To control how the audio is decoded, cast the `audio` column using `dataset.cast_column("audio", datasets.Audio(...))`.
 Useful arguments to `datasets.Audio()` are:
 - `sampling_rate` and `mono=True` to control the sampling rate and number of channels.
-- `decode=False` to skip decoding the audio and just get the MP3 file paths.
+- `decode=False` to skip decoding the audio and just get the MP3 file paths and contents.
+
+The `load_dataset` function also accepts a `columns` parameter, which can be useful for example if you want to skip downloading the audio (see the example below).
 
 ## Running the benchmark
 
@@ -61,14 +175,14 @@ The evaluation is implemented in our [`alt-eval` package](https://github.com/aud
 from datasets import load_dataset
 from alt_eval import compute_metrics
 
-dataset = load_dataset("audioshake/jam-alt", revision="v1.0.0")["test"]
+dataset = load_dataset("audioshake/jam-alt", revision="v1.1.0", split="test")
 # transcriptions: list[str]
 compute_metrics(dataset["text"], transcriptions, languages=dataset["language"])
 ```
 
 For example, the following code can be used to evaluate Whisper:
 ```python
-dataset = load_dataset("audioshake/jam-alt", revision="v1.0.0")["test"]
+dataset = load_dataset("audioshake/jam-alt", revision="v1.1.0", split="test")
 dataset = dataset.cast_column("audio", datasets.Audio(decode=False))  # Get the raw audio file, let Whisper decode it
 
 model = whisper.load_model("tiny")
@@ -78,9 +192,9 @@ transcriptions = [
 ]
 compute_metrics(dataset["text"], transcriptions, languages=dataset["language"])
 ```
-Alternatively, if you already have transcriptions, you might prefer to skip loading the audio:
+Alternatively, if you already have transcriptions, you might prefer to skip loading the `audio` column:
 ```python
-dataset = load_dataset("audioshake/jam-alt", revision="v1.0.0", with_audio=False)["test"]
+dataset = load_dataset("audioshake/jam-alt", revision="v1.1.0", split="test", columns=["name", "text", "language", "license_type"])
 ```
 
 ## Citation
@@ -108,4 +222,4 @@ When using the benchmark, please cite [our paper](https://www.arxiv.org/abs/2408
   address={Rhodes Island, Greece},
   doi={10.1109/ICASSP49357.2023.10096725}
 }
-```
+```

jam-alt.py → loader.py

@@ -11,7 +11,7 @@ from typing import Optional
 import datasets
 
 
-_VERSION = "1.0.0"
+_VERSION = "1.1.0"
 
 
 _CITATION = """\

parquet/all/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed027d2032652b082c79b63008eecea1dacce97400ce1d5cb0326e958302d6e1
+size 409150043

parquet/de/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4fe32692f20daeb06b2b3c253b51dbd5bc4a4c6bb491a64fde821d263d95134
+size 107942102

parquet/en/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c8870334ca9ff4a5166940355169ee36a260725c782a23bd352948d38c83f70
+size 105041371

parquet/es/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96468681e62d4f69729c2d33890bfc17caf9718b4f56cd843ab1468a4c72c8a1
+size 104979012

parquet/fr/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:935eb807ce731d44dfc904a1b8f2dd751b626e8c40774d56ee793b799ec987e0
+size 91218543

push_to_hub.py

@@ -0,0 +1,26 @@
+import argparse
+import datasets
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--repo", type=str, required=True)
+    parser.add_argument("--revision", type=str, required=True)
+    args = parser.parse_args()
+
+    for config_name in ["all", "en", "es", "de", "fr"]:
+        dataset = datasets.load_dataset(
+            "./loader.py", config_name, trust_remote_code=True
+        )
+        dataset.push_to_hub(
+            args.repo,
+            config_name,
+            set_default=(config_name == "all"),
+            data_dir=f"parquet/{config_name}",
+            commit_message=f"Upload parquet/{config_name}",
+            revision=args.revision,
+        )
+
+
+if __name__ == "__main__":
+    main()