Fix speculative decode with full rotating caches#1427
Open
cyq1017 wants to merge 2 commits into
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Speculative decoding can diverge from deterministic non-speculative generation when a
RotatingKVCachereaches the point where it can no longer be trimmed. If a drafted token is rejected after that point, the existing rewind path cannot remove the speculative cache entries.This limits draft length around full rotating caches and falls back to target-only generation when rollback is not safe. Other trimmable cache types keep their existing speculative path.
Fixes #1423.
Tests
uv run --isolated --python 3.12 --with-editable . --with pytest python -m pytest tests/test_generate.py::TestSpeculativeGenerateStep tests/test_generate.py::TestGenerate::test_stream_generate_speculative -qpython3 -m py_compile mlx_lm/generate.py tests/test_generate.pygit diff --check HEAD~2..HEAD