[fix]修复splitpd调度lora报错

vllm/v1/core/sched/scheduler.py

@@ -663,6 +663,9 @@ class Scheduler(SchedulerInterface):
         # For logging.
         scheduled_timestamp = time.monotonic()
 
+        # Record the LoRAs in scheduled_running_reqs
+        scheduled_loras: set[int] = set()
+
         # Use a temporary RequestQueue to collect requests that need to be
         # skipped and put back at the head of the waiting queue later
         skipped_waiting_requests = create_request_queue(self.policy)
@@ -993,8 +996,6 @@ class Scheduler(SchedulerInterface):
                         self.encoder_cache_manager.allocate(request, i)
                     encoder_compute_budget = new_encoder_compute_budget
 
-            # Record the LoRAs in scheduled_running_reqs
-            scheduled_loras: set[int] = set()
         if self.lora_config:
             scheduled_loras = set(
                 req.lora_request.lora_int_id for req in scheduled_running_reqs
@@ -1089,14 +1090,11 @@ class Scheduler(SchedulerInterface):
     
     def schedule(self) -> SchedulerOutput:
         if envs.VLLM_USE_PD_SPLIT: 
-            if self.connector is not None:
-                return self.schedule_default()
-            if self.use_mla:
-                if self.full_cuda_graph and self.num_spec_tokens > 0:
             return self.schedule_split_pd()
         else:
+            if self.connector is not None:
                 return self.schedule_default()
-            else:
+            if self.full_cuda_graph and self.use_mla and self.num_spec_tokens > 0 :
                 return self.schedule_split_pd()
             else:
                 return self.schedule_default()