fix: patch model.forward() directly — survives TRL unwrapping#253
Merged
fix: patch model.forward() directly — survives TRL unwrapping#253
Conversation
TRL unwraps models via Accelerate, stripping wrapper classes. The fix: patch forward() on the model instance itself. This survives unwrapping. - patch_model_for_trl(model) → returns cache_fn - cache_fn(inputs) caches pixel_values from processor output - Patched forward() injects cached pixel_values when TRL omits them - Patched __call__ also injects (covers all call paths) - trl_wrapper passes original model to TRL (not a wrapper) - cache_vision_fn passed through to rollout_func Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
VLMModelWrapper was stripped by TRL/Accelerate's model unwrapping. Fix: patch
model.forward()directly on the instance. Instance method patches survive unwrapping because they're on the model object, not a wrapper.Before:
VLMModelWrapper(model)→ TRL unwraps → wrapper gone → model blindAfter:
patch_model_for_trl(model)→ patchesmodel.forward→ TRL unwraps → patch stays → model sees images🤖 Generated with Claude Code