[Cherry-pick] fix stop seq (#3263)

* fix out-bound value for stop sequence

* catch error if there are out-of-bounds value

* check in offline mode

Author: zoooo0820

fastdeploy/engine/engine.py

@@ -530,6 +530,26 @@ def add_requests(self, task, sampling_params=None, **kwargs):
             llm_logger.error(error_msg)
             raise EngineError(error_msg, error_code=400)
 
+        if request.get("stop_seqs_len") is not None:
+            stop_seqs_len = request.get("stop_seqs_len")
+            max_stop_seqs_num = int(envs.FD_MAX_STOP_SEQS_NUM)
+            if len(stop_seqs_len) > max_stop_seqs_num:
+                error_msg = (
+                    f"Length of stop ({stop_seqs_len}) exceeds the limit max_model_len({max_stop_seqs_num})."
+                    "Please reduce the number of stop or set a lager max_stop_seqs_num by `FD_MAX_STOP_SEQS_NUM`"
+                )
+                llm_logger.error(error_msg)
+                raise EngineError(error_msg, error_code=400)
+            stop_seqs_max_len = int(envs.FD_STOP_SEQS_MAX_LEN)
+            for single_stop_seq_len in stop_seqs_len:
+                if single_stop_seq_len > stop_seqs_max_len:
+                    error_msg = (
+                        f"Length of stop_seqs({single_stop_seq_len}) exceeds the limit max_model_len({stop_seqs_max_len})."
+                        "Please reduce the length of stop sequences or set a larger stop_seqs_max_len by `FD_STOP_SEQS_MAX_LEN`"
+                    )
+                    llm_logger.error(error_msg)
+                    raise EngineError(error_msg, error_code=400)
+
         if self.guided_decoding_checker is not None:
             request, err_msg = self.guided_decoding_checker.schema_format(request)
             if err_msg is not None:

fastdeploy/entrypoints/engine_client.py

@@ -19,6 +19,7 @@
 
 import numpy as np
 
+from fastdeploy import envs
 from fastdeploy.input.preprocess import InputPreprocessor
 from fastdeploy.inter_communicator import IPCSignal, ZmqClient
 from fastdeploy.metrics.work_metrics import work_process_metrics
@@ -144,6 +145,26 @@ def add_requests(self, task):
             api_server_logger.error(error_msg)
             raise EngineError(error_msg, error_code=400)
 
+        if "stop_seqs_len" in task:
+            stop_seqs_len = task["stop_seqs_len"]
+            max_stop_seqs_num = int(envs.FD_MAX_STOP_SEQS_NUM)
+            if len(stop_seqs_len) > max_stop_seqs_num:
+                error_msg = (
+                    f"Length of stop ({stop_seqs_len}) exceeds the limit max_model_len({max_stop_seqs_num})."
+                    "Please reduce the number of stop or set a lager max_stop_seqs_num by `FD_MAX_STOP_SEQS_NUM`"
+                )
+                api_server_logger.error(error_msg)
+                raise EngineError(error_msg, error_code=400)
+            stop_seqs_max_len = int(envs.FD_STOP_SEQS_MAX_LEN)
+            for single_stop_seq_len in stop_seqs_len:
+                if single_stop_seq_len > stop_seqs_max_len:
+                    error_msg = (
+                        f"Length of stop_seqs({single_stop_seq_len}) exceeds the limit max_model_len({stop_seqs_max_len})."
+                        "Please reduce the length of stop sequences or set a larger stop_seqs_max_len by `FD_STOP_SEQS_MAX_LEN`"
+                    )
+                    api_server_logger.error(error_msg)
+                    raise EngineError(error_msg, error_code=400)
+
         task["preprocess_end_time"] = time.time()
         preprocess_cost_time = task["preprocess_end_time"] - task["preprocess_start_time"]
         api_server_logger.info(