fix: streamline model argument handling in server.py

mineru/model/vlm_vllm_model/server.py

@@ -9,26 +9,38 @@ def main():
 
     has_port_arg = False
     has_gpu_memory_utilization_arg = False
-    has_model_arg = False
+    model_path = None
+    model_arg_indices = []
 
+    # 检查现有参数
     for i, arg in enumerate(args):
         if arg == "--port" or arg.startswith("--port="):
             has_port_arg = True
         if arg == "--gpu-memory-utilization" or arg.startswith("--gpu-memory-utilization="):
             has_gpu_memory_utilization_arg = True
-        if arg == "--model" or arg.startswith("--model="):
-            has_model_arg = True
-
+        if arg == "--model":
+            if i + 1 < len(args):
+                model_path = args[i + 1]
+                model_arg_indices.extend([i, i + 1])
+        elif arg.startswith("--model="):
+            model_path = arg.split("=", 1)[1]
+            model_arg_indices.append(i)
+
+    # 从参数列表中移除 --model 参数
+    if model_arg_indices:
+        for index in sorted(model_arg_indices, reverse=True):
+            args.pop(index)
+
+    # 添加默认参数
     if not has_port_arg:
         args.extend(["--port", "30000"])
     if not has_gpu_memory_utilization_arg:
         args.extend(["--gpu-memory-utilization", "0.5"])
-    if not has_model_arg:
-        default_path = auto_download_and_get_model_root_path("/", "vlm")
-        args.extend([default_path])
+    if not model_path:
+        model_path = auto_download_and_get_model_root_path("/", "vlm")
 
-    # 重新构造sys.argv，以便透传所有参数给vllm
-    sys.argv = [sys.argv[0]] + ["serve"] + args
+    # 重构参数，将模型路径作为位置参数
+    sys.argv = [sys.argv[0]] + ["serve", model_path] + args
 
     # 启动vllm服务器
     print(f"start vllm server: {sys.argv}")