A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 22:15 UTC
[Bugfix][API Server] Fix invalid usage of 'ge' and 'le' in port valid… (#13672)
Robin committed
Feb 22, 2025 at 14:05 UTC
c6ed93860fa8338bf7e4dfcce0040046a2f0df48
2 files changed
+12
-1
vllm/entrypoints/api_server.py
+1
-1
@@ -145,7 +145,7 @@ async def run_server(args: Namespace,
145
if __name__ == "__main__":
146
parser = FlexibleArgumentParser()
147
parser.add_argument("--host", type=str, default=None)
148
- parser.add_argument("--port", type=int, default=8000, ge=1024, le=65535)
148
+ parser.add_argument("--port", type=parser.check_port, default=8000)
149
parser.add_argument("--ssl-keyfile", type=str, default=None)
150
parser.add_argument("--ssl-certfile", type=str, default=None)
151
parser.add_argument("--ssl-ca-certs",
vllm/utils.py
+11
@@ -1194,6 +1194,17 @@ class FlexibleArgumentParser(argparse.ArgumentParser):
1194
1195
return super().parse_args(processed_args, namespace)
1196
1197
+ def check_port(self, value):
1198
+ try:
1199
+ value = int(value)
1200
+ except ValueError:
1201
+ raise argparse.ArgumentTypeError("Port must be an integer")
1202
+
1203
+ if not (1024 <= value <= 65535):
1204
+ raise argparse.ArgumentTypeError("Port must be between 1024 and 65535")
1205
+
1206
+ return value
1207
+
1208
def _pull_args_from_config(self, args: List[str]) -> List[str]:
1209
"""Method to pull arguments specified in the config file
1210
into the command-line args variable.