Skip to content

Conversation

rraminen
Copy link

The commits in the PR, pytorch#76141, are required to build DeepSpeed on ROCm without the hipify errors.

cc: @jithunnair-amd

@jithunnair-amd
Copy link
Collaborator

pytorch-ci test builds failed with unrelated error, so @rraminen please let me know if you were able to test these changes locally on a DeepSpeed workload.

@jithunnair-amd
Copy link
Collaborator

retest pytorch please

@rraminen
Copy link
Author

rraminen commented Jun 1, 2022

Tested this PR locally. No issues.

Copy link
Collaborator

@pruthvistony pruthvistony left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pyTorch build is successful.

@jithunnair-amd jithunnair-amd merged commit 80baeab into ROCm:release/1.11 Jun 28, 2022
akashveramd pushed a commit that referenced this pull request Jun 13, 2025
Removes the `numpy` usage and `tolist` CUDA sync when computing
`gatherd_idxs`.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants