fix(qwen3-tts): progressive EOS bias to cut late breath frames#840
Closed
Thorfy wants to merge 1 commit into
Closed
fix(qwen3-tts): progressive EOS bias to cut late breath frames#840Thorfy wants to merge 1 commit into
Thorfy wants to merge 1 commit into
Conversation
After Blaizzy#695 removed the ICL/instruct text-length max_tokens hard-cap, late codec EOS left trailing breath-like frames. Encourage EOS with a progressive logit bias once past max(75, text_len*6), while still honoring caller max_tokens. Co-authored-by: Cursor <cursoragent@cursor.com>
This was referenced Jul 15, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
max_tokenstext-length hard-cap, Qwen3-TTS often keeps sampling a few breath-like codec frames beforecodec_eos.max_tokens(Qwen3-TTS: honor caller-provided max_tokens in instruct and ICL paths. #695), but aftermax(75, text_len * 6)ramp a progressive EOS logit bias so stop fires sooner.Fixes #841
Test plan
python -m unittest mlx_audio.tts.tests.test_qwen3_tts.TestQwen3TTSEosBias TestQwen3TTSMaxTokenstext_len * 6