explicitly move tensors to one device #2054
Merged
Conversation
This file contains 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
* Explicitly move tensors to one device as torch.cat no longer moves tensors (pytorch/pytorch#35045).
Balandat
reviewed
Jul 4, 2022
Balandat
reviewed
Jul 4, 2022
In case the matrix is type CatLazyTensor, the returned matrix/row can be on different devices. So move the delazified object to the the matrix output device.
Balandat
reviewed
Jul 6, 2022
| @@ -76,7 +76,8 @@ def apply_permutation( | |||
| right_permutation = torch.arange(matrix.size(-1), device=matrix.device) | |||
|
|
|||
| # Apply permutations | |||
| return delazify(matrix.__getitem__((*batch_idx, left_permutation.unsqueeze(-1), right_permutation.unsqueeze(-2)))) | |||
| res = delazify(matrix.__getitem__((*batch_idx, left_permutation.unsqueeze(-1), right_permutation.unsqueeze(-2)))) | |||
There was a problem hiding this comment.
this line is too long also
There was a problem hiding this comment.
Sorry about that, still getting used to pull requests and pre-commit. Don't think the problem was with this line in permutation.py (this one should be 118 long). I think the issue was with my first commit because I resolved the changes, but did not actually push a fix.
fixed line too long flake8 error
|
Hmm I am not sure about the "build" checks - seems like these aren't actually running? Are they some duplicates of the "run test suite" test? I can't merge w/o those and I'm not sure how to run them. @gpleiss maybe you have an idea? |
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Outdated suggestions cannot be applied.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.
Explicitly move tensors to one device before calling torch.cat as torch.cat no longer moves tensors silently (pytorch/pytorch#35407). This solves one of the issues in #2053 resulting when the partition size is set to 0.