[MUSA][9/N] Add FA3 attention backend support through MATE (MUSA AI Tensor Engine) (#22051)

Co-authored-by: zhiguo.qin <zhiguo.qin@mthreads.com>
This commit is contained in:
ori
2026-04-10 14:18:39 -07:00
committed by GitHub
co-authored by zhiguo.qin
parent 6af34b95b6
commit f7a1740101
8 changed files with 958 additions and 11 deletions
+5 -1
View File
@@ -53,6 +53,7 @@ from sglang.srt.utils.common import (
is_hip,
is_hopper_with_cuda_12_3,
is_mps,
is_musa,
is_no_spec_infer_or_topk_one,
is_npu,
is_remote_url,
@@ -2584,7 +2585,10 @@ class ServerArgs:
def _handle_page_size(self):
if self.page_size is None:
self.page_size = 1
if not is_musa():
self.page_size = 1
else:
self.page_size = 64
def _handle_amd_specifics(self):
if is_hip():