Project structure of: james4ever0/whisper

• whisper Open-source speech recognition library and tools
  • CHANGELOG.md Latest openai/whisper project updates, fixes, and features.
  • data
    • README.md Convert, preprocess, timings, labels, multilingual datasets
  • model-card.md Whisper AI: Speech Recognition, Translation, Challenges
  • notebooks Notebooks: Contains scripts for ASR, WER calculation, and multilingual tasks.
    • LibriSpeech.py Initialize LibriSpeech, load test-clean split, perform inference, calculate WER.
    • Multilingual_ASR.py Multilingual ASR/MT script, performs package installs and language model training.
  • pyproject.toml Python code formatting settings
  • README.md Whisper: Multitask Transformer for diverse language speech recognition.
  • requirements.txt Python packages and Triton version requirements.
  • setup.py Set up Python package with requirements.
  • tests Test audio, normalization, timing, and transcription functionality.
    • conftest.py Imports libraries, sets seed.
    • test_audio.py Test audio loading, spectrogram calculation, and consistency.
    • test_normalizer.py Tests number, date, and text normalization classes.
    • test_timing.py Median filter performance test on CPU and GPU.
    • test_tokenizer.py Test tokenizer functionality across languages.
    • test_transcribe.py Test Whisper AI transcription functionality.
  • whisper Whisper: Library for audio-to-text, efficient language detection and transcription.
    • __init__.py Whisper: Library for model download, verification, and availability.
    • __main__.py Transcribes audio using Whisper CLI.
    • audio.py Audio preprocessing for Tensor or Torch tensors.
    • decoding.py Whisper model, efficient audio language detection.
    • model.py Audio-to-text model: encoder-decoder, convolutional layers, attention
    • normalizers Text normalizers for multiple languages.
      • __init__.py Text normalizer initialization.
      • basic.py Normalizes text, removes symbols and diacritics.
      • english.py English language number and text normalizer
    • timing.py Median filtering, dynamic time warping, text-to-speech alignment.
    • tokenizer.py Efficient text tokenization with Tiktoken library support.
    • transcribe.py Speech transcription with Whisper model, customizable.
    • triton_ops.py Triton parallel Dynamic Time Warp, CUDA Median Filter
    • utils.py Utility, compression ratio, timestamp formatting. WriteTSV, JSON.
    • version.py Sets version of "whisper" to 20231117