feat[vLLM]: Make VoxtralProcessor compatible with the Transformers modeling backend for vLLM - #47574
Closed
harshaljanjani wants to merge 1 commit into
Closed
Conversation
4 tasks
Contributor
|
[For maintainers] Suggested jobs to run (before merge) run-slow: voxtral |
Contributor
CI recapDashboard: View test results in Grafana |
eustlb
reviewed
Jul 27, 2026
eustlb
left a comment
Collaborator
There was a problem hiding this comment.
Thanks a lot for starting this! Nevertheless it's better here to have apply_chat_template rely on __call__ rather than two diff paths (not your at all fault though, was implement like that from the beginning).
Since this is tricky, I am opening a PR to supersede. Will add you as a coauthor there of course! 🤗
Contributor
Author
|
Thanks for the attribution and makes sense, closing! |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What does this PR do?
→ A blocker to the companion vLLM PR for Voxtral audio generation through the Transformers backend.
→
VoxtralProcessor.__call__rejects audio-token text and routes audio only throughapply_chat_template. Makes__call__acceptaudiolike the other audio processors; sinceMistralCommonBackenddoes not encode[AUDIO]from text, splices the audio token id into the encoded segments. No modeling changes needed.→ Verified no regressions on the Transformers-side in
tests/models/voxtral/(143 non-slow pass; the 5 failing slow integration tests fail identically onmainlocally, exact-match generation drift, unrelated to this change).→ I've tried to explain why each non-trivial change was made in a one-line comment given much of this involved unwrapping failure modes layer by layer and uncovering deeper issues in the call chain :)
cc: @eustlb @ebezzam @vasqu
Before (vLLM Voxtral Generation)
After (vLLM Voxtral Generation, After Both PRs)
Code Agent Policy
Before submitting