Fix AttributeError in RayCollisionFunction.backward#182
Open
Mr-Neutr0n wants to merge 1 commit intoopenai:mainfrom
Open
Fix AttributeError in RayCollisionFunction.backward#182Mr-Neutr0n wants to merge 1 commit intoopenai:mainfrom
Mr-Neutr0n wants to merge 1 commit intoopenai:mainfrom
Conversation
RayCollisionFunction.forward correctly saves tensors via ctx.save_for_backward(), but backward() retrieves them using the non-existent attribute ctx.input_tensors instead of the correct ctx.saved_tensors. This causes an AttributeError whenever gradients are backpropagated through checkpointed ray casting (i.e., when cast_rays is called with checkpoint=True).
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
RayCollisionFunction.backwardusesctx.input_tensorsto retrieve tensors saved during the forward pass, butctx.input_tensorsdoes not exist in the PyTorchautograd.FunctionAPI. The correct attribute isctx.saved_tensors(populated byctx.save_for_backward()inforward).AttributeErrorwhenever gradients flow through checkpointed ray casting (cast_rayswithcheckpoint=True), which is the default path whenray_batch_size < total_rays.Fix
Replace
ctx.input_tensorswithctx.saved_tensorson line 114 ofshap_e/rendering/raycast/cast.py.Test plan
forwardusesctx.save_for_backward(origins, directions, faces, vertices)backwardnow unpacks viactx.saved_tensorsto matchcheckpoint=Trueand confirm gradients propagate without error