diff --git a/vllm_ascend/models/deepseek_mtp.py b/vllm_ascend/models/deepseek_mtp.py
index b3daa6cc4..0c4f17394 100644
--- a/vllm_ascend/models/deepseek_mtp.py
+++ b/vllm_ascend/models/deepseek_mtp.py
@@ -166,7 +166,7 @@ class CustomDeepSeekMultiTokenPredictor(DeepSeekMultiTokenPredictor):
     def compute_logits(
         self,
         hidden_states: torch.Tensor,
-        sampling_metadata,  # type: ignore
+        sampling_metadata=None,  # type: ignore
         spec_step_idx: int = 0,
     ) -> torch.Tensor:
         current_step_idx = (spec_step_idx % self.num_mtp_layers)
diff --git a/vllm_ascend/models/qwen3_next.py b/vllm_ascend/models/qwen3_next.py
index 7d1481e9a..175a529ab 100644
--- a/vllm_ascend/models/qwen3_next.py
+++ b/vllm_ascend/models/qwen3_next.py
@@ -986,7 +986,7 @@ class Qwen3NextForCausalLM(nn.Module, HasInnerState, SupportsLoRA, SupportsPP,
     def compute_logits(
             self,
             hidden_states: torch.Tensor,
-            sampling_metadata,  # type: ignore
+            sampling_metadata=None,  # type: ignore
     ) -> Optional[torch.Tensor]:
         return self.logits_processor(self.lm_head, hidden_states,
                                      sampling_metadata)
diff --git a/vllm_ascend/torchair/models/qwen2.py b/vllm_ascend/torchair/models/qwen2.py
index 56620dc8a..6e4990d7b 100644
--- a/vllm_ascend/torchair/models/qwen2.py
+++ b/vllm_ascend/torchair/models/qwen2.py
@@ -344,7 +344,7 @@ class CustomQwen2ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
     def compute_logits(
             self,
             hidden_states: torch.Tensor,
-            sampling_metadata,  # type: ignore
+            sampling_metadata=None,  # type: ignore
     ) -> Optional[torch.Tensor]:
         logits = self.logits_processor(self.lm_head, hidden_states,
                                        sampling_metadata)
diff --git a/vllm_ascend/torchair/models/torchair_deepseek_mtp.py b/vllm_ascend/torchair/models/torchair_deepseek_mtp.py
index 6ef90f46e..a7c5a6ee4 100644
--- a/vllm_ascend/torchair/models/torchair_deepseek_mtp.py
+++ b/vllm_ascend/torchair/models/torchair_deepseek_mtp.py
@@ -170,7 +170,7 @@ class TorchairDeepSeekMultiTokenPredictor(DeepSeekMultiTokenPredictor):
     def compute_logits(
         self,
         hidden_states: torch.Tensor,
-        sampling_metadata,  # type: ignore
+        sampling_metadata=None,  # type: ignore
         spec_step_idx: int = 0,
     ) -> torch.Tensor:
         current_step_idx = (spec_step_idx % self.num_mtp_layers)
diff --git a/vllm_ascend/torchair/models/torchair_pangu_moe.py b/vllm_ascend/torchair/models/torchair_pangu_moe.py
index e38dc789d..195ffdedd 100644
--- a/vllm_ascend/torchair/models/torchair_pangu_moe.py
+++ b/vllm_ascend/torchair/models/torchair_pangu_moe.py
@@ -936,7 +936,7 @@ class PanguProMoEForCausalLM(nn.Module, SupportsPP):
     def compute_logits(
             self,
             hidden_states: torch.Tensor,
-            sampling_metadata,  # type: ignore
+            sampling_metadata=None,  # type: ignore
     ) -> Optional[torch.Tensor]:
         logits = self.logits_processor(self.lm_head, hidden_states,
                                        sampling_metadata)
diff --git a/vllm_ascend/worker/model_runner_v1.py b/vllm_ascend/worker/model_runner_v1.py
index 98aeac6da..f4656dddf 100644
--- a/vllm_ascend/worker/model_runner_v1.py
+++ b/vllm_ascend/worker/model_runner_v1.py
@@ -2784,9 +2784,10 @@ class NPUModelRunner(LoRAModelRunnerMixin):
             for idx in range(len(kv_cache_tensor.shared_by)):
                 layer_name = kv_cache_tensor.shared_by[idx]
                 if "linear_attn" in layer_name:
+                    # for mamba linear attention
                     for layer_name_inner in kv_cache_tensor.shared_by:
-                        if "self_attn" in layer_name_inner or layer_name_inner in kv_cache_raw_tensors.keys(
-                        ):
+                        if ("attn" in layer_name_inner and "linear_attn" not in layer_name_inner) or \
+                            layer_name_inner in kv_cache_raw_tensors.keys():
                             continue
                         if self.vllm_config.kv_transfer_config is None:
                             tensor = torch.zeros(kv_cache_tensor.size,
@@ -2800,7 +2801,8 @@ class NPUModelRunner(LoRAModelRunnerMixin):
                             tensor = self._align_memory(
                                 tensor, alignment)[:kv_cache_tensor.size]
                         kv_cache_raw_tensors[layer_name_inner] = tensor
-                elif "self_attn" in layer_name:
+                elif "attn" in layer_name:
+                    # for other attentions, e.g., self_attn, sliding window attn
                     if self.vllm_config.kv_transfer_config is None:
                         k_tensor = torch.zeros(kv_cache_tensor.size // 2,
                                                dtype=torch.int8,