fix: Float8Tensor as_strided workaround#503
Open
livepeer-robot wants to merge 1 commit intodaydreamlive:mainfrom
Open
fix: Float8Tensor as_strided workaround#503livepeer-robot wants to merge 1 commit intodaydreamlive:mainfrom
livepeer-robot wants to merge 1 commit intodaydreamlive:mainfrom
Conversation
Float8Tensor (torchao) does not implement aten.as_strided which is used internally by unflatten/view/reshape. Add safe_unflatten helper that detects Float8Tensor and casts to float32 before the operation. Applied to causal_vace_model.py where the error occurs in DenoiseBlock. Fixes daydreamlive#501
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Float8Tensor (torchao) does not implement
aten.as_stridedwhich is used internally byunflatten/view/reshape. This adds asafe_unflattenhelper that detects Float8Tensor and casts to float32 before the operation.Applied to
causal_vace_model.pywhere the error occurs in DenoiseBlock.Fixes #501