Added missing gpu support for relu op and also disabled padding for gemms#2806
Open
pemeliya wants to merge 2 commits intor2.17-rocm-enhancedfrom
Open
Added missing gpu support for relu op and also disabled padding for gemms#2806pemeliya wants to merge 2 commits intor2.17-rocm-enhancedfrom
pemeliya wants to merge 2 commits intor2.17-rocm-enhancedfrom
Conversation
i-chaochen
reviewed
Jan 17, 2025
| pre_pipeline.AddPass<CublasPadForGemms>(rocm_compute_capability, | ||
| req.data_type, req.multiple_of); | ||
| } | ||
| // for (const auto& req : HipblasPaddingRequirements) { |
Collaborator
There was a problem hiding this comment.
Could you add comment here to indicate this is CUDA passs so we disable it.
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Note that CublasPadForGemms was only enabled for the parity with NV, but in fact it does not bring any better performance on ROCM platform