Put row lengths on the same device on gpu#113
Merged
oliverholworthy merged 3 commits intoNVIDIA-Merlin:mainfrom Mar 22, 2023
Merged
Put row lengths on the same device on gpu#113oliverholworthy merged 3 commits intoNVIDIA-Merlin:mainfrom
oliverholworthy merged 3 commits intoNVIDIA-Merlin:mainfrom
Conversation
| if len(row_lengths.shape) == 2: | ||
| zero_value = zero_value.view(-1, 1) | ||
| return torch.cat((zero_value, torch.cumsum(row_lengths, 0))) | ||
| return torch.cat((zero_value, torch.cumsum(row_lengths, 0).to(device=self.device))) |
Contributor
There was a problem hiding this comment.
assuming torch.cat preserves the device of the values being concatented, maybe an alternative to this would be to change the line where zero_value is defined to use row_lengths.device instead of self.device?
Contributor
There was a problem hiding this comment.
that might make the row_lengths to offsets method more robust, but fact that the error is showing up suggests that even with this fix we might end up with a mismatch between the loader.device attribute and the output tensor device?
Contributor
Author
There was a problem hiding this comment.
Thanks for the suggestion. Honestly I'm not sure which option is best, but it also works, and I think your suggestion is better. Changed in de593b2.
oliverholworthy
approved these changes
Mar 22, 2023
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
In a multi-gpu setting, tensors may be generated on different devices. This PR forces
torch.cumsum(row_lengths, 0)to be on the same device as thezero_valuetensor. If they are on different devices,torch.cat()can't concatenate them, e.g.,