[Pallas] Remove torch.empty in tracing#6897
Merged
alanwaketan merged 4 commits intomasterfrom Apr 8, 2024
Merged
Conversation
Collaborator
Author
|
This PR is ready for review. |
Collaborator
|
I will take a look |
JackCaoG
approved these changes
Apr 8, 2024
Collaborator
|
Do we need this in 2.3 release? I don't consider this as the critical fix, wdyt? |
Collaborator
Author
I think it is, but I will hold it until I finished some e2e modeling tests. Thanks for the review. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary:
Previously we rely on torch.empty to create some empty tensors as the outputs from the Pallas and then make Pallas as in-place ops. However, it turns out that torch.empty will actually allocate memory and therefore it's expansive to use. In this change, I switched to simply pass the shapes and dtypes to construct the graph node.
Test Plan:
PJRT_DEVICE=TPU python test/test_pallas.py
Performance benchmarks can be found: http://shortn/_wdmom7I6q7