fix lint

Signed-off-by: whx-sjtu <[email protected]>

Author: whx-sjtu

vllm_ascend/attention/utils.py

@@ -8,10 +8,10 @@
                                           is_v1_kv_transfer_group)
 from vllm.forward_context import ForwardContext, get_forward_context
 
-# We find that _npu_paged_attention still performes better than
+# We find that _npu_paged_attention still performs better than
 # npu_fused_infer_attention_score in some cases. We allow to execute
 # _npu_paged_attention in this cases. This should be removed once
-# npu_fused_infer_attention_score performes better on all scenarios.
+# npu_fused_infer_attention_score performs better on all scenarios.
 PAGED_ATTENTION_LIST = [1, 2, 3, 4]
 
 

vllm_ascend/compilation/acl_graph.py

@@ -509,7 +509,7 @@ def set_graph_params(aclgraph_capture_sizes: set[int]):
     )
 
 
-def update_graph_params_workspaces(num_tokens: int, workspace: int):
+def update_graph_params_workspaces(num_tokens: int, workspace: torch.Tensor):
     global _graph_params
     if _graph_params is not None:
         _graph_params.workspaces[num_tokens] = weak_ref_tensors(workspace)