Add audio search

.gitignore

@@ -0,0 +1,153 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/

app.py

@@ -1,4 +1,4 @@
-from transformers import ClapModel, ClapProcessor
+from transformers import ClapModel, ClapProcessor, AutoFeatureExtractor
 import gradio as gr
 import torch
 import torchaudio
@@ -9,7 +9,8 @@ from qdrant_client.http.models import Distance, VectorParams
 from qdrant_client.http import models
 
 
-
+import dotenv
+dotenv.load_dotenv()
 
 
 class ClapSSGradio():
@@ -25,7 +26,7 @@ class ClapSSGradio():
 
         self.model = ClapModel.from_pretrained(
             f"Audiogen/{name}", use_auth_token=os.getenv('HUGGINGFACE_API_TOKEN'))
-        self.tokenizer = ClapProcessor.from_pretrained(
+        self.processor = ClapProcessor.from_pretrained(
             f"Audiogen/{name}", use_auth_token=os.getenv('HUGGINGFACE_API_TOKEN'))
 
         self.sas_token = os.environ['AZURE_SAS_TOKEN']
@@ -40,15 +41,41 @@ class ClapSSGradio():
         # print(self.client.get_collection(collection_name=self.name))
 
     @torch.no_grad()
-    def _embed_query(self, query):
-        inputs = self.tokenizer(
-            query, return_tensors="pt", padding='max_length', max_length=77, truncation=True)
-        return self.model.get_text_features(**inputs).cpu().numpy().tolist()[0]
-
-    def _similarity_search(self, query, threshold):
+    def _embed_query(self, query, audio_file):
+        if audio_file is not None:
+            waveform, sample_rate = torchaudio.load(audio_file.name)
+            print("Waveform shape:", waveform.shape)
+            waveform = torchaudio.functional.resample(
+                waveform, sample_rate, 48000)
+            print("Resampled waveform shape:", waveform.shape)
+
+            if waveform.shape[-1] < 480000:
+                waveform = torch.nn.functional.pad(
+                    waveform, (0, 48000 - waveform.shape[-1]))
+            elif waveform.shape[-1] > 480000:
+                waveform = waveform[..., :480000]
+
+            audio_prompt_features = self.processor(
+                audios=waveform.mean(0), return_tensors='pt', sampling_rate=48000
+            )['input_features']
+            print("Audio prompt features shape:", audio_prompt_features.shape)
+            e = self.model.get_audio_features(
+                input_features=audio_prompt_features)[0]
+
+            if any(torch.isnan(e)):
+                raise ValueError("Audio features are NaN")
+            print("Embeddings: ", e.shape)
+            return e
+        else:
+            inputs = self.processor(
+                query, return_tensors="pt", padding='max_length', max_length=77, truncation=True)
+
+            return self.model.get_text_features(**inputs).cpu().numpy().tolist()[0]
+
+    def _similarity_search(self, query, threshold, audio_file):
         results = self.client.search(
             collection_name=self.name,
-            query_vector=self._embed_query(query),
+            query_vector=self._embed_query(query, audio_file),
             limit=self.k,
             score_threshold=threshold,
         )
@@ -94,14 +121,19 @@ class ClapSSGradio():
             with gr.Row():
                 with gr.Column(variant='panel'):
                     search = gr.Textbox(placeholder='Search Samples')
-                    float_input = gr.Number(label='Similarity threshold [min: 0.1 max: 1]', default=0.5, minimum=0.1, maximum=1)
+                    float_input = gr.Number(
+                        label='Similarity threshold [min: 0.1 max: 1]', value=0.5, minimum=0.1, maximum=1)
+                    audio_file = gr.File(
+                        label='Upload an Audio File', type="file")
+                    search_button = gr.Button("Search", label='Search')
                 with gr.Column():
                     audioboxes = []
                     gr.Markdown("Output")
                     for i in range(self.k):
                         t = gr.components.Audio(label=f"{i}", visible=True)
                         audioboxes.append(t)
-            search.submit(fn=self._similarity_search, inputs=[search, float_input], outputs=audioboxes)
+            search_button.click(fn=self._similarity_search, inputs=[
+                search, float_input, audio_file], outputs=audioboxes)
             ui.launch(share=share)