TensorRT-LLM backend based on WhisperS2T#290

Merged

info-wordcab merged 32 commits intomainfrom 285-try-whispers2t-as-potential-backend

Apr 1, 2024

+54,837-501

Commits on Mar 25, 2024

Commits on Mar 29, 2024

Commits on Mar 30, 2024

Added download/convert option for distilled Whisper models
Aleks
committed
Adjusted max text token lengths, and updated jax/flax requirements
Aleks
committed
Adjusted max text token lengths, and updated jax/flax requirements
Aleks
committed
Fixing align model for n_mels=128 models
Aleks
committed
Switched default model to distil-large-v3
Aleks
committed
Removed sha256 for whisper model downloads
Aleks
committed
Switched to safetensors, fixed n_mels issue for distil-large-v3
Aleks
committed
Removing jax/flax
Aleks
committed
Adjusted n_mels change for v3 models only
Aleks
committed
Testing latest faster-whisper library
Aleks
committed
Added diarization backend options
Aleks
committed
Moved convert ctranslate2 script to new scripts folder
Aleks
committed
Added Cython dependency, rounding timestamps to 2 decimal places
Aleks
committed
Added Cython dependency, rounding timestamps to 2 decimal places
Aleks
committed
Added Cython/youtokentome dependencies
Aleks
committed

Commits on Mar 31, 2024

Enabling longform diarization, skeleton for punct-based alignment
Aleks
committed

Commits on Apr 1, 2024

Adding README for self-hosted models
Aleks
committed
Documentation update in .env file
Aleks
committed
Experimenting with larger batch size
Aleks
committed
Experimenting with larger batch size
Aleks
committed
Adding batch_size option to transcription call
Aleks
committed
Updating tensorrt_llm dependency
Aleks
committed
Updating tensorrt_llm dependency
Aleks
committed
Updating numpy dependency
Aleks
committed
Updating numpy dependency
Aleks
committed
Updating numpy dependency
Aleks
committed
Updating numpy dependency
Aleks
committed