fix: 修复deepseek在PP>1开chunkprefill精度问题

vllm/v1/worker/gpu_model_runner.py

@@ -1638,6 +1638,9 @@ class GPUModelRunnerBase(LoRAModelRunnerMixin):
         # Mask out the sampled tokens that should not be sampled.
         for i in discard_sampled_tokens_req_indices:
             valid_sampled_token_ids[i].clear()
+        if spec_token_ids is not None:
+            for i in discard_sampled_tokens_req_indices:
+                spec_token_ids[i].clear()
 
         # Cache the sampled tokens in the model runner, so that the scheduler
         # doesn't need to send them back.
@@ -3519,6 +3522,9 @@ class GPUModelRunnerMTP(GPUModelRunnerBase):
         # Mask out the sampled tokens that should not be sampled.
         for i in discard_sampled_tokens_req_indices:
             valid_sampled_token_ids[i].clear()
+        if spec_token_ids is not None:
+            for i in discard_sampled_tokens_req_indices:
+                spec_token_ids[i].clear()
 
         # Cache the sampled tokens in the model runner, so that the scheduler
         # doesn't need to send them back.