Expose MLLM MTP draft counters#504
Open
Thump604 wants to merge 1 commit intowaybarrios:mainfrom
Open
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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\n- track whether MLLM MTP actually attempted drafts on each primary response\n- carry attempted and accepted draft counters through MLLMScheduler outputs\n- expose those counters on BatchedEngine GenerationOutput for MLLM routes\n\n## Why\nThe MLLM MTP wrapper could previously make it hard to distinguish an installed MTP path from a step that actually drafted. That can make skipped speculative steps look like zero-acceptance MTP attempts. This PR records attempts only after the draft path actually runs, and counts accepted draft responses separately.\n\n## Tests\n- AI_RUNTIME_BYPASS_SAFETY_GATE=1 /opt/ai-runtime/venv-live/bin/python -m pytest -q tests/test_mllm_mtp_routing.py -q\n- uvx ruff check vllm_mlx/mllm_batch_generator.py vllm_mlx/mllm_scheduler.py vllm_mlx/engine/batched.py vllm_mlx/engine/base.py vllm_mlx/request.py tests/test_mllm_mtp_routing.py --select E,F,W --ignore E402,E501,E731,F811,F841\n- black --check --target-version py312 vllm_mlx/mllm_batch_generator.py vllm_mlx/mllm_scheduler.py vllm_mlx/engine/batched.py vllm_mlx/engine/base.py vllm_mlx/request.py tests/test_mllm_mtp_routing.py