Repository navigation
[CUDA] Fix gather_mm with sorted_indices=True - #4771
Open
luciusacarr wants to merge 2 commits into
Open
luciusacarr wants to merge 2 commits into
luciusacarr wants to merge 2 commits into
Conversation
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 #4727
With
sorted_indices=True,gather_mmreturned the wrong rows when xs size-1 axis came fromexpand_dimsor x[:, None, :]. This is becauseensure_batch_contiguousreturnedx.strides(-2)for row contiguous inputs. On a size-1 axis, that stride is arbitrary. By returningx.shape(-1)` instead, we read the real row step for row contiguous data. A test is added to avoid future regressions.On my 5090 on CUDA 13.3, this fix made the number of wrong rows go from 60/64 to 0/64.