Fix modular pipeline stateful cache reset and TaylorSeer/FBC context crash#27
Draft
cursor[bot] wants to merge 1 commit into
Draft
Fix modular pipeline stateful cache reset and TaylorSeer/FBC context crash#27cursor[bot] wants to merge 1 commit into
cursor[bot] wants to merge 1 commit into
Conversation
ModularPipeline.__call__ now resets stateful inference caches on all components after each run, matching DiffusionPipeline.maybe_free_model_hooks. Without this, PAB/FasterCache/MagCache hook state from run N corrupts run N+1. TaylorSeer and FirstBlockCache hooks auto-set an inference context when none is provided, consistent with MagCache, so enable_cache() does not crash on denoise paths that omit cache_context. Co-authored-by: Simon Lynch <srlynch1@users.noreply.github.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Fixes two cache-related correctness bugs:
Modular pipelines never reset stateful caches between runs —
ModularPipeline.__call__now calls_reset_stateful_caches()on all components after each invocation (in afinallyblock), matching the cache-reset behavior inDiffusionPipeline.maybe_free_model_hooks(). Without this, inference caches (PAB, FasterCache, MagCache, etc.) retain staleiteration/step_index/cached tensors from run N and corrupt run N+1.TaylorSeer and FirstBlockCache crash without
cache_context— Both hooks now auto-set an"inference"context when none is provided, consistent with MagCache. PreviouslyStateManager.get_state()raisedValueErroron the first forward when pipelines omittedcache_context.Validation
pytest tests/modular_pipelines/test_modular_pipeline_cache_reset.py tests/hooks/test_taylorseer_cache_context.py(3 tests, all pass)utils/check_copies.pychange: invoke ruff viapython -m ruffso copy-check hooks work whenruffis not on PATH.