[causal attn mask] replaced for loop over tensor with pytorch tensor ops #208
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
dabeschte
changed the title
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Outdated suggestions cannot be applied.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.
The original causal attention mask generation is very slow, especially when the tensor is created on the GPU, because it needs to make 1000s of calls.
I tried to compile it...which works and makes it fast too, but compilation unfortunately also takes a long time when using a long sequence length.
This implementation is ~20-70x faster depending on the sequence lengths and since it is re-created for every SDPA, this accumulates to multiple seconds per step for larger videos