Fix GenerationMixin compat with transformers>=5.14.0 - #1200
Merged
rwightman merged 1 commit intoJul 31, 2026
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #1199
Summary
transformers5.14.0 added a call toself.get_experts_implementation()inside GenerationMixin._optimize_model_for_decode(), invoked unconditionally fromgenerate()'s decode loop. That method is only defined ontransformers.PreTrainedModel, not onGenerationMixinitself.MultimodalGenerationWrapper(src/open_clip/generation.py) only inheritsnn.Module+GenerationMixinby design -- it's a thin wrapper around open_clip's own generative models, not a fullPreTrainedModel-- so it breaks:AttributeError: 'MultimodalGenerationWrapper' object has no attribute 'get_experts_implementation'`. Full analysis and bisection in #1199.Fix: a trivial override. This wrapper never has MoE submodels, so there's nothing to report -- an empty dict makes the caller's
"grouped_mm" in experts_implementation.values()check a no-op, restoring correct behavior.Test plan
tests/test_coca2.py(27 tests, includingtest_modern_coca_generateand alltest_coca_generate_*) -- pass againsttransformers==5.14.1(the actual broken version) with this fix, unpinned.tests/test_naflex_mammut.py,tests/test_mammut.py-- allgenerate-related tests across MaMMUT/NaFlex variants also pass.