[feat]支持mtp模型full_cuda_graph

vllm/v1/spec_decode/eagle.py

@@ -232,7 +232,7 @@ class EagleProposer:
         sample_hidden_states = last_hidden_states[last_token_indices]
         logits = self.model.compute_logits(sample_hidden_states, None)
 
-        draft_token_ids = torch.argmax(logits, dim=-1)
+        draft_token_ids = logits.argmax(dim=-1)
 
         # Early exit if there is only one draft token to be generated.
         if self.num_speculative_tokens == 1:
@@ -380,7 +380,7 @@ class EagleProposer:
             logits = self.model.compute_logits(last_hidden_states[:batch_size],
                                                None)
 
-            # # TODO(wenlong): get more than one token for tree attention
+            # TODO(wenlong): get more than one token for tree attention
             draft_token_ids = logits.argmax(dim=-1)
             draft_token_ids_list.append(draft_token_ids)