[perf]Support MOE Multi-stream in Deepseek (#947)

### What this PR does / why we need it? Support MOE inner Multi-stream for Deepseek. This feature requires graph mode with mc2 enabled. --------- Signed-off-by: David9857 <985700846@qq.com>
2025-06-05 23:39:38 +08:00
parent 908a851a77
commit 78431b3469
6 changed files with 133 additions and 45 deletions
--- a/tests/singlecard/test_ascend_config.py
+++ b/tests/singlecard/test_ascend_config.py
@@ -114,5 +114,6 @@ def test_ascend_config_load_error():
            },
        }
        with VllmRunner("facebook/opt-125m",
+                        enforce_eager=False,
                        additional_config=input_additional_config_fake_2):
            pass