[BugFix] race condition [is_fetching] causing multiple fetch requests (PaddlePaddle#5238)

* RouterArgs port str -> int

* fix race condition [is_fetching] causing multiple fetch requests

* bugfix: Delete duplicate input_ids tensor creation

Author: ST-XX

fastdeploy/engine/common_engine.py

@@ -680,7 +680,6 @@ def _schedule_request_to_worker_v1(self):
         def _fetch_request():
             try:
                 nonlocal is_fetching
-                is_fetching = True
                 num_prefill_batch = min(
                     int(self.resource_manager.available_batch()),
                     self.cfg.max_prefill_batch,
@@ -803,6 +802,7 @@ def _fetch_request():
                     continue
                 if self.cfg.scheduler_config.splitwise_role != "mixed":
                     if not is_fetching:
+                        is_fetching = True
                         get_request_pool.submit(_fetch_request)
 
                 else:
@@ -813,6 +813,7 @@ def _fetch_request():
                     ):
                         # Check if the thread pool is still available to avoid submitting tasks to a shutdown thread pool.
                         try:
+                            is_fetching = True
                             get_request_pool.submit(_fetch_request)
                         except RuntimeError as e:
                             if "shutdown" in str(e):

fastdeploy/engine/sched/resource_manager_v1.py

@@ -414,7 +414,6 @@ def _get_num_new_tokens(self, request, token_budget):
         ):
             input_ids_lst = request.prompt_token_ids + request.output_token_ids
             input_ids = paddle.to_tensor(input_ids_lst, dtype="int64")
-            input_ids = paddle.to_tensor(input_ids_lst, dtype="int64")
             image_patch_id = inputs["image_patch_id"]
 
             if request.multimodal_img_boundaries is None:

fastdeploy/router/router.py

@@ -32,7 +32,7 @@ class RouterArgs:
     """
     Host address to bind the router server
     """
-    port: str = "9000"
+    port: int = 9000
     """
     Port to bind the router server.
     """
@@ -55,7 +55,7 @@ def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
         )
         parser.add_argument(
             "--port",
-            type=str,
+            type=int,
             default=RouterArgs.port,
             help="Port number to bind the router server",
         )