[Misc] Move --grpc CLI argument into make_arg_parser (#38570)

Signed-off-by: Chang Su <chang.s.su@oracle.com>

vllm/entrypoints/cli/serve.py

@@ -136,13 +136,6 @@ class ServeSubcommand(CLISubcommand):
         )
 
         serve_parser = make_arg_parser(serve_parser)
-        serve_parser.add_argument(
-            "--grpc",
-            action="store_true",
-            default=False,
-            help="Launch a gRPC server instead of the HTTP OpenAI-compatible "
-            "server. Requires: pip install vllm[grpc].",
-        )
         serve_parser.epilog = VLLM_SUBCMD_PARSER_EPILOG.format(subcmd=self.name)
         return serve_parser
 

vllm/entrypoints/openai/cli_args.py

@@ -345,6 +345,13 @@ def make_arg_parser(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
         "Must be a YAML with the following options: "
         "https://docs.vllm.ai/en/latest/configuration/serve_args.html",
     )
+    parser.add_argument(
+        "--grpc",
+        action="store_true",
+        default=False,
+        help="Launch a gRPC server instead of the HTTP OpenAI-compatible "
+        "server. Requires: pip install vllm[grpc].",
+    )
     parser = FrontendArgs.add_cli_args(parser)
     parser = AsyncEngineArgs.add_cli_args(parser)