fix lint

Signed-off-by: whx-sjtu <2952154980@qq.com>

Author: whx-sjtu

tests/e2e/singlecard/test_multistream_overlap_shared_expert.py

@@ -52,7 +52,8 @@ def test_models_with_multistream_overlap_shared_expert(
                 "multistream_overlap_shared_expert": True,
             },
     ) as runner:
-        vllm_moe_ms_eager_outputs = runner.model.generate(prompts, sampling_params)
+        vllm_moe_ms_eager_outputs = runner.model.generate(
+            prompts, sampling_params)
 
     with VllmRunner(
             model,
@@ -62,7 +63,8 @@ def test_models_with_multistream_overlap_shared_expert(
                 "multistream_overlap_shared_expert": True,
             },
     ) as runner:
-        vllm_moe_ms_aclgraph_outputs = runner.model.generate(prompts, sampling_params)
+        vllm_moe_ms_aclgraph_outputs = runner.model.generate(
+            prompts, sampling_params)
 
     with VllmRunner(
             model,

vllm_ascend/ops/common_fused_moe.py

@@ -439,7 +439,8 @@ def forward(
     ) -> tuple[torch.Tensor, torch.Tensor]:
         # Make sure the shared experts stream begins after hidden_states are ready.
         if self.multistream_overlap_shared_expert:
-            self.shared_expert_stream.wait_stream(torch.npu.current_stream())
+            self.shared_expert_stream.wait_stream(  # type: ignore
+                torch.npu.current_stream())
         with npu_stream_switch(self.shared_expert_stream,
                                enabled=self.multistream_overlap_shared_expert):
             # Use a separate stream to run shared experts.

vllm_ascend/torchair/ops/torchair_fused_moe.py

@@ -1254,7 +1254,8 @@ def forward(self,
             log2phy=self.log2phy,
             global_redundant_expert_num=self.global_redundant_expert_num,
             shared_experts=shared_experts if self.torchair_graph_enabled
-            and self.multistream_overlap_shared_expert and not is_prefill else None,
+            and self.multistream_overlap_shared_expert and not is_prefill else
+            None,
             mc2_mask=mc2_mask,
             quantized_x_for_share=quantized_x_for_share,
             dynamic_scale_for_share=dynamic_scale_for_share,

vllm_ascend/utils.py

@@ -322,7 +322,8 @@ def update_aclgraph_sizes(vllm_config: VllmConfig) -> None:
         # TODO: Find out whether we need to take into account the pp_size
         parallel_factor = 1 + num_comm_groups + int(
             parallel_config.enable_expert_parallel) + int(
-            vllm_config.additional_config.get("multistream_overlap_shared_expert", False))
+                vllm_config.additional_config.get(
+                    "multistream_overlap_shared_expert", False))
         if is_moe_model(vllm_config):
             parallel_factor += (parallel_config.data_parallel_size > 1)
         # Calculate maximum supported batch sizes considering model architecture on the A2 Hardware Device