feat: Qwen3-ASR stateful CoreML decoder conversion#18
Draft
Alex-Wengg wants to merge 4 commits intomainfrom
Draft
feat: Qwen3-ASR stateful CoreML decoder conversion#18Alex-Wengg wants to merge 4 commits intomainfrom
Alex-Wengg wants to merge 4 commits intomainfrom
Conversation
Stateful decoder with GPU-resident KV cache for Qwen3-ASR-0.6B. Includes conversion script and mlpackage output.
- convert-qwen3-asr.py: main CLI for exporting all components (audio_encoder, embedding, lm_head, decoder variants) - convert_decoder_fused.py: fused decoder with lmHead for faster inference - individual_components.py: wrapper modules for each component - pyproject.toml: uv dependencies - README.md: basic usage
- Stateful CoreML decoder with GPU-resident KV cache (1.1x → 2.9x RTFx) - Fused lmHead into decoder (~8ms/tok savings) - Benchmark warm-up for consistent timing - Int8 quantization revisited (899 MB, same quality) - WER outlier analysis (proper noun limitations) - MLX comparison benchmarks
Alex-Wengg
force-pushed
the
feature/qwen3-asr-coreml
branch
from
db0bc45 to
42ea7c9
Compare
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
convert_stateful_decoder.pyfor converting Qwen3-ASR-0.6B decoder to a stateful CoreML model with GPU-resident KV cacheqwen3_asr_decoder_stateful.mlpackage(LFS-tracked)QWEN3_ASR_COREML.mdreference doc covering architecture analysis, conversion approach, and integration notesTest plan
qwen3-asrbranch).mlmodelcfrom HuggingFace repo matches local conversion🤖 Generated with Claude Code