fix test cases failed on 5/30 in nightly pipeline (#26775)

python/sglang/test/ascend/test_ascend_utils.py

@@ -168,7 +168,7 @@ GLM_4_5V_WEIGHTS_PATH = os.path.join(MODEL_WEIGHTS_DIR, "ZhipuAI/GLM-4.5V")
 JANUS_PRO_1B_WEIGHTS_PATH = os.path.join(MODEL_WEIGHTS_DIR, "deepseek-ai/Janus-Pro-1B")
 JANUS_PRO_7B_WEIGHTS_PATH = os.path.join(MODEL_WEIGHTS_DIR, "deepseek-ai/Janus-Pro-7B")
 KIMI_VL_A3B_INSTRUCT_WEIGHTS_PATH = os.path.join(
-    MODEL_WEIGHTS_DIR, "Kimi/Kimi-VL-A3B-Instruct"
+    MODEL_WEIGHTS_DIR, "moonshotai/Kimi-VL-A3B-Instruct"
 )
 LLAMA_3_2_11B_VISION_INSTRUCT_WEIGHTS_PATH = os.path.join(
     MODEL_WEIGHTS_DIR, "LLM-Research/Llama-3.2-11B-Vision-Instruct"
@@ -436,6 +436,7 @@ def get_benchmark_args(
         gsp_num_turns=gsp_num_turns,
         header=header,
         max_concurrency=max_concurrency,
+        ready_check_timeout_sec=0,
     )
 
 

test/registered/ascend/basic_function/parallel_strategy/expert_parallelism/test_npu_deepep_auto_qwen3_next.py

@@ -81,6 +81,7 @@ class TestQwen3Next(CustomTestCase):
                 "HCCL_ALGO": "level0:NA;level1:ring",
                 "SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK": "20",
                 "HCCL_BUFFSIZE": "2000",
+                "GDN_ATTN_BACKEND_TRITON": "1",
                 **os.environ,
             },
         )

test/registered/ascend/basic_function/parallel_strategy/expert_parallelism/test_npu_deepep_low_latency_qwen3_next.py

@@ -85,6 +85,7 @@ class TestQwen3Next(CustomTestCase):
                 "HCCL_ALGO": "level0:NA;level1:ring",
                 "SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK": "160",
                 "HCCL_BUFFSIZE": "2048",
+                "GDN_ATTN_BACKEND_TRITON": "1",
                 **os.environ,
             },
         )

test/registered/ascend/basic_function/parameter/test_npu_no_overlap_scheduler.py

@@ -1,8 +1,7 @@
 import unittest
 
-from sglang.test.ascend.test_ascend_utils import run_mmlu_test
 from sglang.test.ci.ci_register import register_npu_ci
-from sglang.test.test_utils import CustomTestCase
+from sglang.test.test_utils import CustomTestCase, run_mmlu_test
 
 register_npu_ci(
     est_time=400,