[Pallas] Allow setting FlashAttention's causal mask #6792
Merged
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
JackCaoG
reviewed
Mar 21, 2024
| v = torch.randn(3, 2, 128, 4).to("xla") | ||
|
|
||
| # The causal mask is turned on by default in the wrapper. | ||
| # It masks out the top right triangle of the attention matrix, therefore it speeds up the compute but also changes the output. |
There was a problem hiding this comment.
what do you mean by also changes the output?
There was a problem hiding this comment.
The output will be different from the eager attention output, which is reflected in the test.
JackCaoG
approved these changes
Mar 21, 2024
|
Thanks Jack for approving. |
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Outdated suggestions cannot be applied.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.
Summary:
This pull request channels the causal mask to our wrapper.
Test Plan:
PJRT_DEVICE=TPU python test/test_pallas.py -v -k test_flash_attention_wrapper_causal