Spaces:
Running
Running
Commit History
27 days ago
-
Update VoxPreference Running
Verified signaturef7fba16 -
Update VoxPreference
Verified signature7a48bd2 -
Update VoxPreference
Verified signature11ca35b -
Update VoxPreference
Verified signature8776f20 -
Update VoxPreference
Verified signature757ecb4 -
Update VoxPreference
Verified signaturec4ad58f -
Update VoxPreference
Verified signature20fb2bc -
Update VoxPreference
Verified signature91ff1f3 -
Update VoxPreference
Verified signature4cccf8f -
Update VoxPreference
Verified signature08c216f -
Preference Grammar pipeline: one image for frontend and backend
Verified signature9b3bb37
-
Add HEAD method for root endpoint in app.py
- Introduced a new HEAD method for the root endpoint to allow clients to check the availability of the API without retrieving the full response. - Enhanced logging to track calls to the new HEAD endpoint, maintaining consistency with existing logging practices.ebdd03f -
Remove training and validation CSV files from the data directory
- Deleted `train.csv` and `val.csv` files to clean up the dataset as they are no longer needed. - Updated the `split_shuffle_data.py` script to ensure it correctly handles the absence of these files during data processing.b1eab91
-
Refactor IPA conversion logic in text_to_ipa.py for improved efficiency
- Enhanced the convert_to_ipa function to handle empty and whitespace-only inputs gracefully. - Implemented multiprocessing to optimize IPA conversion for larger texts by processing words in batches. - Improved error handling to provide detailed feedback for failed conversions. - Updated the final IPA output to concatenate results from individual word conversions, ensuring accuracy.0845174 -
Enhance Wav2Vec2 fine-tuning process in finetune_model.py
- Refactored data loading to utilize pandas for CSV input, improving dataset handling. - Introduced a comprehensive MetricsTracker class for detailed metrics logging and visualization during training. - Implemented a custom LabelSmoothingCTCLoss for improved loss calculation. - Enhanced the Trainer class to include metrics tracking and early stopping based on WER. - Added visualization capabilities for training metrics and detailed analysis. - Updated training configuration for better performance and model selection based on WER. - Improved requirements.txt to include additional dependencies for metrics tracking and evaluation.8d221d1
-
Refactor transcription logic in app.py to segment words
- Updated the transcription process to split the output into words instead of sentences, improving the granularity of the segments. - Adjusted the calculation of start and end times to reflect word durations, enhancing the accuracy of the transcription output.83294f4
-
Refactor transcription logic in app.py to segment sentences
- Updated the transcription process to split the output into sentences instead of words, improving the granularity of the segments. - Adjusted the calculation of start and end times to reflect sentence durations, enhancing the accuracy of the transcription output. - Incorporated regular expressions for better sentence detection and handling.54d1e1d -
Enhance transcription functionality in app.py
- Updated the transcription endpoint to calculate and return word segments with their respective start and end times. - Added a new endpoint for converting individual words to IPA, improving the overall functionality of the application. - Enhanced error handling and logging for both transcription and IPA conversion processes.8a07ad8
-
Add espeak installation to Dockerfile
- Updated Dockerfile to install espeak for text-to-speech functionality. - Included cleanup command to reduce image size by removing apt cache.4a743d7
-
Update README.md with comprehensive project documentation
- Expanded README to include detailed project overview, features, architecture, and usage instructions. - Added sections for API reference, environment variables, logging, cache management, development setup, finetuning instructions, troubleshooting, license, and acknowledgements. - Enhanced clarity and organization for better user guidance and onboarding.d94733d -
Refactor caching configuration and improve error logging in app.py
- Updated environment variables to use HF_HOME and NUMBA_CACHE_DIR for caching. - Enhanced error logging in the transcription process for better debugging.11dce03 -
Add logging functionality to app.py for request tracking and error handling
- Integrated logging configuration to capture application events. - Added middleware to log incoming requests and responses. - Enhanced endpoint logging for root, health, and transcription routes. - Implemented error logging during the transcription process.a7c12f5 -
Set up Dockerfile and app.py for caching Transformers models
dd0ebb6 -
Implement Wav2Vec2 fine-tuning for Nigerian English speech recognition
- Added finetune_model.py for training Wav2Vec2 model with custom dataset. - Integrated audio processing and text cleaning functions. - Created a data collator for CTC with padding. - Configured training arguments and initiated training process. - Added manual evaluation and model saving after training. Enhance requirements for audio processing and model training - Updated requirements.txt to include necessary libraries for training and audio processing. Add IPA conversion utility - Introduced text_to_ipa.py for converting text to IPA using phonemizer. Implement audio statistics retrieval - Created get_audio_stats.py to compute total duration and size of audio files categorized by gender. Develop data splitting and shuffling script - Added split_shuffle_data.py to load, shuffle, and split audio data into training and validation sets.e203eab -
initial commit
Verified signaturef2076f5