Add fused kernel for Hawk forward and backward #1
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
…r perf at low BS, don't recompute log(2*exp) 3 times per iteration
|
LGTM 👍 |
Thanks Stefan, appreciate your detailed review. Merging now :) |
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Outdated suggestions cannot be applied.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.
This PR adds a fused kernel scan kernel for the forward and backward pass. It is around 2x faster than the Triton kernel I originally wrote. It should also use less memory, due to the activation recomputation in the backward pass (benchmarks pending).
Benchmarks
Also benchmarked against
torch.nn.functional.scaled_dot_product_attentionwithhd=64Forward Pass
bs=8,d_model=1024Forward + Backward
bs=8,d_model=1024