Merge branch 'v0.7.2-dev-wm' into 'v0.7.2-dev'

[fix]修复fused_moe.py中fused_moe接口未初始化moe_ep_size导致的deekseek等模型报错

See merge request dcutoolkit/deeplearing/vllm!89

vllm/model_executor/layers/fused_moe/fused_moe.py

@@ -1672,7 +1672,7 @@ def fused_moe(
     a2_scale: Optional[torch.Tensor] = None,
     block_shape: Optional[List[int]] = None,
     use_nn_moe: Optional[bool] = False,
-    moe_ep_size: Optional[int] = None,
+    moe_ep_size: Optional[int] = 1,
     start_expert: Optional[int] = None,
     end_expert: Optional[int] = None,
 ) -> torch.Tensor: