Open
Conversation
Summary: Add TritonBench operator to benchmark the flex attention forward pass inductor kernel (triton_tem_fused_flex_attention_0). Compares aten (eager flex_attention) vs inductor (torch.compile) with causal mask, sweeping seq_len from 128 to 16384. Reports latency, speedup, and tflops (adjusted for block sparsity). Default config: B=8, H=16, D=128, bf16. Reviewed By: stashuk-olek Differential Revision: D95461825
Summary: Add TritonBench operator to benchmark the flex attention backward pass inductor kernel (triton_tem_fused_flex_attention_backward_zeros_1). Uses FWD_ONLY=True but manually times backward via output.backward(dy, retain_graph=True). Compares aten (eager) vs inductor (torch.compile). Backward FLOP count uses 2.5x multiplier (2.0 bwd + 0.5 recompute). Default config: B=8, H=16, D=128, bf16, requires_grad=True on q/k/v. Reviewed By: stashuk-olek Differential Revision: D95461827
OmarPavel
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
OmarPavel
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
OmarPavel
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
|
@OmarPavel has exported this pull request. If you are a Meta employee, you can view the originating Diff in D95461827. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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:
Add TritonBench operator to benchmark the flex attention backward pass
inductor kernel (triton_tem_fused_flex_attention_backward_zeros_1).
Uses FWD_ONLY=True but manually times backward via
output.backward(dy, retain_graph=True). Compares aten (eager) vs
inductor (torch.compile). Backward FLOP count uses 2.5x multiplier
(2.0 bwd + 0.5 recompute).
Default config: B=8, H=16, D=128, bf16, requires_grad=True on q/k/v.
Reviewed By: stashuk-olek
Differential Revision: D95461827