Merge branch 'v0.9.2-dev-wm-1127' into 'v0.9.2-dev'

解决宽松mtp request_ids报错找不到问题

See merge request dcutoolkit/deeplearing/vllm!276

vllm/v1/spec_decode/utils.py

@@ -52,7 +52,7 @@ class DraftProbs(ABC):  # type: ignore[call-arg]
     draft_probs: torch.Tensor
 
     # The request id list.
-    _req_ids: list[str]
+    _req_ids: list[str] = []
 
     def __init__(self, draft_probs, req_ids):
         assert len(req_ids) == len(draft_probs)
@@ -64,10 +64,15 @@ class DraftProbs(ABC):  # type: ignore[call-arg]
                tmp_req_ids: list[str]):
         diff_req_ids = [item for item in self._req_ids if item not in tmp_req_ids]
         index = [self._req_ids.index(req_id) for req_id in diff_req_ids]
-        self._req_ids = diff_req_ids
-        self.draft_probs = self.draft_probs[index]
+        index_tensor = async_tensor_h2d(
+                    index,
+                    dtype=torch.int32,
+                    target_device=self.draft_probs.device,
+                    pin_memory=True)
+        self.draft_probs = self.draft_probs[index_tensor]
         self.draft_probs = torch.cat([self.draft_probs, draft_probs])
         
+        self._req_ids = diff_req_ids
         self._req_ids.extend(tmp_req_ids)
         assert len(self._req_ids) == len(self.draft_probs)
 
@@ -76,7 +81,12 @@ class DraftProbs(ABC):  # type: ignore[call-arg]
         if new_req_ids != self._req_ids:
             # Batch contents changed - prune removed sequences.
             index = [self._req_ids.index(req_id) for req_id in new_req_ids]
-            self.draft_probs = self.draft_probs[index]
+            index_tensor = async_tensor_h2d(
+                    index,
+                    dtype=torch.int32,
+                    target_device=self.draft_probs.device,
+                    pin_memory=True)
+            self.draft_probs = self.draft_probs[index_tensor]
             self._req_ids = new_req_ids
 
     def get_probs(self, req_ids: list[str]):

vllm/v1/worker/gpu_model_runner.py

@@ -1531,6 +1531,8 @@ class GPUModelRunnerBase(LoRAModelRunnerMixin):
                 )
                 sampler_output.sampled_token_ids = output_token_ids
             else:
+                sampling_metadata.all_greedy = True
+                sampling_metadata.all_random = False
                 sampler_output = self.sampler(
                     logits=logits,
                     sampling_metadata=sampling_metadata,
@@ -1643,12 +1645,7 @@ class GPUModelRunnerBase(LoRAModelRunnerMixin):
             if not envs.VLLM_REJECT_SAMPLE_OPT:
                 spec_token_ids = spec_result
             else:
-                spec_token_ids, draft_probs = spec_result
-                if self.draft_probs is None:
-                    self.draft_probs = DraftProbs(
-                    draft_probs, self.input_batch.req_ids)
-                else:
-                    self.draft_probs.update(draft_probs, self.input_batch.req_ids)
+                spec_token_ids, _ = spec_result
 
         # Clear KVConnector state after all KVs are generated.
         if has_kv_transfer_group():
@@ -1787,8 +1784,14 @@ class GPUModelRunnerBase(LoRAModelRunnerMixin):
                 draft_token_ids = draft_result
                 spec_token_ids = draft_token_ids.tolist()
                 return spec_token_ids
-            
+
+            draft_req_ids = list(scheduler_output.num_scheduled_tokens.keys())
             draft_token_ids, draft_probs = draft_result
+            if self.draft_probs is None:
+                self.draft_probs = DraftProbs(
+                    draft_probs, draft_req_ids)
+            else:
+                self.draft_probs.update(draft_probs, draft_req_ids)
             spec_token_ids = draft_token_ids.tolist()
             return spec_token_ids, draft_probs
 
@@ -3357,6 +3360,8 @@ class GPUModelRunnerMTP(GPUModelRunnerBase):
                 )
                 sampler_output.sampled_token_ids = output_token_ids
             else:
+                sampling_metadata.all_greedy = True
+                sampling_metadata.all_random = False
                 sampler_output = self.sampler(
                     logits=logits,
                     sampling_metadata=sampling_metadata,
@@ -3438,12 +3443,8 @@ class GPUModelRunnerMTP(GPUModelRunnerBase):
             if not envs.VLLM_REJECT_SAMPLE_OPT:
                 spec_token_ids = spec_result
             else:
-                spec_token_ids, draft_probs = spec_result
-                if self.draft_probs is None:
-                    self.draft_probs = DraftProbs(
-                    draft_probs, self.input_batch.req_ids)
-                else:
-                    self.draft_probs.update(draft_probs, self.input_batch.req_ids)
+                spec_token_ids, _ = spec_result
+                
         if max_gen_len == 1:
             # No spec decode tokens.
             valid_sampled_token_ids = sampled_token_ids.tolist()
@@ -3607,7 +3608,13 @@ class GPUModelRunnerMTP(GPUModelRunnerBase):
                 spec_token_ids = draft_token_ids.tolist()
                 return spec_token_ids
             
+            draft_req_ids = list(scheduler_output.num_scheduled_tokens.keys())
             draft_token_ids, draft_probs = draft_result
+            if self.draft_probs is None:
+                self.draft_probs = DraftProbs(
+                    draft_probs, draft_req_ids)
+            else:
+                self.draft_probs.update(draft_probs, draft_req_ids)
             spec_token_ids = draft_token_ids.tolist()
             return spec_token_ids, draft_probs
         return spec_token_ids