[Bugfix][Frontend] Run the serve arg checks for vllm launch too (#52825)

Signed-off-by: Vineeth Sai <vineethsai4444@gmail.com>
This commit is contained in:
Vineeth Sai Varikuntla
2026-08-19 04:48:50 +00:00
committed by GitHub
parent 08afae2786
commit 86a89a9348
2 changed files with 47 additions and 1 deletions
+3 -1
View File
@@ -412,7 +412,9 @@ def make_arg_parser(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
def validate_parsed_serve_args(args: argparse.Namespace):
"""Quick checks for model serve args that raise prior to loading."""
if hasattr(args, "subparser") and args.subparser != "serve":
# `vllm launch <component>` builds its parser with make_arg_parser too (see
# LaunchSubcommandBase.add_cli_args), so its args are serve args as well.
if hasattr(args, "subparser") and args.subparser not in ("serve", "launch"):
return
# Ensure that the chat template is valid; raises if it likely isn't