[Bugfix] Fix Qwen3/DSV3/DSV3.2 model support (#11510)

2025-10-16 15:14:09 +08:00
parent b0d20cdec7
commit 3cceaa381a
12 changed files with 102 additions and 33 deletions
--- a/test/srt/ascend/test_ascend_deepep.py
+++ b/test/srt/ascend/test_ascend_deepep.py
@@ -38,13 +38,10 @@ class TestAscendDeepEP(CustomTestCase):
            "--quantization",
            "w8a8_int8",
            "--mem-fraction-static",
-            0.9,
-            "--max-running-requests",
-            32,
+            0.8,
            "--disable-radix-cache",
            "--chunked-prefill-size",
            32768,
-            "--disable-cuda-graph",
            "--tp-size",
            16,
            "--dp-size",
@@ -58,7 +55,7 @@ class TestAscendDeepEP(CustomTestCase):
        ]

        cls.extra_envs = {
-            "HCCL_BUFFSIZE": "500",
+            "HCCL_BUFFSIZE": "1000",
            "SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK": "32",
            "SGLANG_NPU_USE_MLAPO": "1",
        }