[Fix] Fix the output of scatter on MLU device #1881
Merged
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Outdated suggestions cannot be applied.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.
Motivation
MMDP or MMDDP use origin scatter_gather to scatter Tensor to one or more device(s).
And the result of scatter_map should be a List of Tensor, each item is for one device.
However, the modified scatter_gather for MLUDP or MLUDDP only returns the pure Tensor.
It causes dimension errors when scatter this result again.
Modification
Change the output of scatter_map, thus the output is just the same size as GPU or CPU result.
BC-breaking (Optional)
No BC-breaking.
Checklist
Before PR:
After PR: