Merge remote-tracking branch 'origin/main'

Author: yangshihao6

.github/workflows/vllm_ascend_test_pr_light.yaml

@@ -135,10 +135,12 @@ jobs:
           pytest -sv --cov --cov-report=xml:unittests-coverage.xml tests/ut \
             --ignore tests/ut/torchair/models/test_torchair_deepseek_mtp.py \
             --ignore tests/ut/torchair/models/test_torchair_deepseek_v2.py \
-            --ignore tests/ut/models/test_qwen2_vl.py \
-            --ignore tests/ut/models/test_qwen2_5_vl.py \
-            --ignore tests/ut/models/test_qwen2_5_vl_without_padding.py \
-            --ignore tests/ut/model_loder
+            --ignore tests/ut/model_loader/netloader/test_netloader_elastic.py \
+            --ignore tests/ut/kv_connector/test_remote_prefill_lifecycle.py \
+            --ignore tests/ut/kv_connector/test_remote_decode_lifecycle.py \
+            --ignore tests/ut/kv_connector/test_llmdatadist_connector.py \
+            --ignore tests/ut/ops/test_linear.py \
+            --ignore tests/ut/core/test_scheduler_dynamic_batch.py
 
       - name: Upload coverage to Codecov
         # only upload coverage when commits merged

tests/e2e/nightly/ops/test_fused_moe.py

@@ -28,7 +28,8 @@
 import torch_npu
 from vllm.model_executor.layers.activation import SiluAndMul
 
-from vllm_ascend.ops.fused_moe.experts_selector import select_experts
+from vllm_ascend.ops.fused_moe.experts_selector import (
+    check_npu_moe_gating_top_k, select_experts)
 from vllm_ascend.ops.fused_moe.moe_mlp import unified_apply_mlp
 from vllm_ascend.ops.fused_moe.token_dispatcher import \
     TokenDispatcherWithAllGather
@@ -303,7 +304,10 @@ def test_select_experts(
             e_score_correction_bias=e_score_correction_bias,
         )
 
-        if use_grouped_topk:
+        call_moe_gatingtopk = check_npu_moe_gating_top_k(
+            hidden_states, topk, topk_group, num_expert_group, scoring_func,
+            custom_routing_function)
+        if not call_moe_gatingtopk and use_grouped_topk:
             mock_native_grouped_topk.assert_called_once()
         else:
             mock_native_grouped_topk.assert_not_called()