[fix]修复缺少参数等错误

vllm/model_executor/layers/fused_moe/batched_deep_gemm_moe.py

@@ -590,6 +590,7 @@ class BatchedDeepGemmExperts(mk.FusedMoEPermuteExpertsUnpermute):
         expert_tokens_meta: mk.ExpertTokensMetadata | None,
         apply_router_weight_on_input: bool,
         use_nn_moe: bool | None = False,
+        **_
     ):
         assert expert_tokens_meta is not None
         expert_num_tokens = expert_tokens_meta.expert_num_tokens

vllm/v1/attention/backend.py

@@ -342,7 +342,7 @@ class CommonAttentionMetadata:
     block_table_tensor: torch.Tensor
     slot_mapping: torch.Tensor
 
-    num_kv_actual_tokens: int
+    num_kv_actual_tokens: int | None = None
     
     seq_indexes_list: list[int] | None = None
     scatter_indexes_tensor: torch.Tensor | None = None
@@ -434,6 +434,7 @@ class CommonAttentionMetadata:
             else None,
             num_reqs=num_actual_reqs,
             num_actual_tokens=num_actual_tokens,
+            num_kv_actual_tokens=num_actual_tokens,
             max_query_len=self.max_query_len,
             max_seq_len=self.max_seq_len,
             block_table_tensor=self.block_table_tensor[:num_actual_reqs],