Skip to content

[Fix] Set fa3 as default MHA backend on Hopper #17425

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Fridge003 merged 2 commits into from
Jan 21, 2026
Merged

[Fix] Set fa3 as default MHA backend on Hopper #17425

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0122c1a
fix
Fridge003 Jan 20, 2026
351f58a
upd
Fridge003 Jan 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 4 additions & 6 deletions python/sglang/srt/server_args.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,6 @@
get_quantization_config,
is_blackwell_supported,
is_cuda,
is_fa3_default_architecture,
is_flashinfer_available,
is_hip,
is_hopper_with_cuda_12_3,
Expand Down Expand Up @@ -1628,11 +1627,10 @@ def _handle_attention_backend_compatibility(self):

if not use_mla_backend:
# MHA architecture
if (
is_hopper_with_cuda_12_3()
and is_no_spec_infer_or_topk_one(self)
and is_fa3_default_architecture(self.model_config.hf_config)
):
if is_hopper_with_cuda_12_3() and is_no_spec_infer_or_topk_one(self):
# Note: flashinfer 0.6.1 caused performance regression on Hopper attention kernel
# Before the kernel is fixed, we choose fa3 as the default backend on Hopper MHA
# ref: https://github.com/sgl-project/sglang/issues/17411
self.attention_backend = "fa3"
elif (
is_sm100_supported()
Expand Down
Loading