[BUGFIX] clear request (#4320)

Co-authored-by: ltd0924 <[email protected]>

Author: ltd0924

fastdeploy/engine/common_engine.py

@@ -806,7 +806,8 @@ def clear_data(self):
             llm_logger.info("Clear Data: Start")
             self.token_processor.clear_data()
             self.engine_worker_queue.clear_data()
-            self.zmq_server.req_dict.clear()
+            self.send_response_server.req_dict.clear()
+            self.recv_request_server.req_dict.clear()
             llm_logger.info("Clear Data: Successfully")
             return True
         except Exception as e:

fastdeploy/entrypoints/openai/api_server.py

@@ -499,6 +499,7 @@ def control_scheduler(request: ControlSchedulerRequest):
         return JSONResponse(content=content.model_dump(), status_code=500)
 
     if request.reset:
+        llm_engine.engine.clear_data()
         llm_engine.engine.scheduler.reset()
 
     if request.load_shards_num or request.reallocate_shard:

fastdeploy/rl/dynamic_weight_manager.py

@@ -223,6 +223,7 @@ def check_model_weights_status(model_weights_status, model_runner, pid):
         while model_weights_status.value[0] != ModelWeightsStatus.NORMAL:
             if model_weights_status.value[0] == ModelWeightsStatus.UPDATING:
                 logger.info("infer engine stopped! start to load new checkpoint...")
+                model_runner.clear_requests()
                 model_runner.update_parameters(pid)
                 while model_weights_status.value[0] != ModelWeightsStatus.NORMAL:
                     time.sleep(0.01)

fastdeploy/worker/worker_process.py

@@ -337,6 +337,8 @@ def event_loop_normal(self) -> None:
                         self.worker.model_runner,
                         self.parallel_config.engine_worker_queue_port,
                     )
+                    logger.info(f"current task queue data: {self.task_queue.num_tasks()}")
+                    self.task_queue.clear_data()
                     self.model_weights_signal[0] = ModelWeightsStatus.NORMAL
                     logger.info(f"Rank: {self.local_rank} has updated or cleared parameters.")