benchmark decoding attention kernel with cudnn (#2467)

Co-authored-by: root <[email protected]>
sgl-project · Dec 17, 2024 · e210266 · e210266
1 parent bd61961
commit e210266
Show file tree

Hide file tree

Showing 2 changed files with 405 additions and 172 deletions.
diff --git a/benchmark/kernels/decoding_attention_triton/sglang_triton_vs_flashinfer.py b/benchmark/kernels/decoding_attention_triton/sglang_triton_vs_flashinfer.py