Add comprehensive coverage of non-transformers ASR implementations #218
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- Update Common Voice dataset from v13 to v17 (latest available) - Update language count from 108 to 124 languages in Common Voice 17 - Update all dataset URLs and references throughout chapter5 files - Update Whisper model reference from whisper-large-v2 to whisper-large-v3 - Update training examples and code snippets to use latest dataset version - Maintain educational content structure while using current resources Files updated: - chapters/en/chapter5/choosing_dataset.mdx - chapters/en/chapter5/evaluation.mdx - chapters/en/chapter5/fine-tuning.mdx - chapters/en/chapter5/asr_models.mdx - chapters/en/_toctree.yml (minor formatting fix)
- Add detailed section on Moonshine ASR: edge-optimized, 5x faster for short audio - Add detailed section on Kyutai STT: real-time streaming capabilities - Include architecture comparison table with performance characteristics - Add code examples for using Moonshine and Kyutai models - Update model selection table with new ASR alternatives - Add model-specific dataset recommendations in choosing_dataset.mdx - Provide guidance on when to choose each model architecture - Update summary to reflect expanded ASR landscape This addresses the Whisper-centric nature of Chapter 5 by providing comprehensive coverage of modern ASR alternatives with different optimization focuses.
- Create new alternative_implementations.mdx with extensive coverage of: * whisper.cpp: C++ port, 10x faster CPU inference, edge deployment * faster-whisper: 4x faster with CTranslate2, GPU support, streaming * insanely-fast-whisper: 36x faster, ultra-low latency, real-time focus * MLX-Whisper: Apple Silicon optimized, 50% faster on M-series chips * Lightning-Whisper-MLX: 10x faster MLX implementation * WhisperKit: On-device Apple deployment with Swift APIs * Conformer models: Edge computing, 5.26x realtime on wearables * Squeezeformer: Memory-efficient architecture alternative - Add comprehensive performance comparison table with metrics for: * Speed vs original implementation * Memory usage characteristics * Platform-specific optimizations * Accuracy trade-offs * Deployment recommendations - Include practical code examples for each implementation - Add deployment strategies for edge, mobile, and real-time scenarios - Provide benchmarking utilities and best practices - Add integration examples with existing transformers workflows - Update _toctree.yml to include new file in navigation - Add cross-references in asr_models.mdx and choosing_dataset.mdx - Maintain educational structure while expanding ecosystem coverage This addresses the need for comprehensive coverage of ASR implementations beyond the transformers library, providing practical guidance for production deployments across different platforms and use cases.
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This PR adds comprehensive coverage of ASR (mainly whisper) implementations beyond the transformers library, addressing the need for practical guidance on production deployments across different platforms and use cases.
This new unit mentions:
A Comprehensive Code Examples
A Practical Deployment Guidance