Pin all layout by default and use all_reduce based all_gather when la… #3568
Conversation
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
hjm-aws
reviewed
May 16, 2022
hjm-aws
reviewed
May 16, 2022
|
@hjm-aws If this looks OK to you I will merge it today. |
hjm-aws
approved these changes
May 18, 2022
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Outdated suggestions cannot be applied.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.
…yout is pin
In PyTorch/XLA we need to pin all layouts for cross-core-communciation ops since program is built separately for every core. However there is not an easy way to pin the layout for
all-gatheron TPU. In this pr I changed theall-gatheris to useall-reducewhen user actually wants to pin the layout. I also changed thelayout-pinningtoTrueby defualt. It was set to False because I can't pinall-gather.@hjm-aws I am guessing you guys still prefer to use
all_gatherdirectly so I set thepin_layoutto False for all xla_process_group. Let me know if that works for you.FYI @hjm-aws @ronghanghu