[Ascend] optimize Qwen-vl on Ascend (#10556)

Co-authored-by: wangqihui01 <wangqh10@163.com>
This commit is contained in:
ronnie_zheng
2025-09-22 17:18:16 -07:00
committed by GitHub
co-authored by wangqihui01
parent d27a6f7092
commit 095093ee5a
5 changed files with 99 additions and 14 deletions
+1 -1
View File
@@ -1840,7 +1840,7 @@ class ServerArgs:
parser.add_argument(
"--mm-attention-backend",
type=str,
choices=["sdpa", "fa3", "triton_attn"],
choices=["sdpa", "fa3", "triton_attn", "ascend_attn"],
default=ServerArgs.mm_attention_backend,
help="Set multimodal attention backend.",
)