#sdy Export pass to make function inputs/outputs evenly sharded. #42
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
![copybara-service[bot]](https://avatars.githubusercontent.com/in/44061?s=60&v=4){kind=small}
copybara-service
bot
force-pushed
the
test_659602009
branch
2 times, most recently
from
3a52334 to
1dbff3f
Compare
copybara-service
bot
force-pushed
the
test_659602009
branch
from
1dbff3f to
5acf234
Compare
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Outdated suggestions cannot be applied.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.
#sdy Export pass to make function inputs/outputs evenly sharded.
After propagation, function inputs/outputs may require padding. We can't allow this. GSPMD makes the tensor fully replicated if one dimension requires padding, however we can improve on that. We find the best sub-sharding per dimension by taking the largest dimension sharding prefix that would only require basic communication - a dynamic-slice for the input tensor, or a basic all-gather on the output tensor.
Note that as part of the work to add explicit reshard ops, we need to make it add reshards for these inputs/outputs, so the explicit reshard pass should come after this pass.