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 13, 03:15 UTC
[Bugfix] Fix host and port join for ipv6 in bench serve (#28679)
Signed-off-by: Scott Zhang <scottzh@fb.com> Co-authored-by: Scott Zhang <scottzh@fb.com>
scottzh8 committed
Nov 16, 2025 at 02:20 UTC
3bc11757984ce256905d1b8517d50b514af8b175
1 file changed
+4
-2
vllm/benchmarks/serve.py
+4
-2
@@ -49,6 +49,7 @@ from vllm.benchmarks.lib.ready_checker import wait_for_endpoint
49
from vllm.benchmarks.lib.utils import convert_to_pytorch_benchmark_format, write_to_json
50
from vllm.transformers_utils.tokenizer import get_tokenizer
51
from vllm.utils.gc_utils import freeze_gc_heap
52
+from vllm.utils.network_utils import join_host_port
53
54
MILLISECONDS_TO_SECONDS_CONVERSION = 1000
55
@@ -1333,8 +1334,9 @@ async def main_async(args: argparse.Namespace) -> dict[str, Any]:
1334
api_url = f"{args.base_url}{args.endpoint}"
1335
base_url = f"{args.base_url}"
1336
else:
1336
- api_url = f"http://{args.host}:{args.port}{args.endpoint}"
1337
- base_url = f"http://{args.host}:{args.port}"
1337
+ host_port = join_host_port(args.host, args.port)
1338
+ api_url = f"http://{host_port}{args.endpoint}"
1339
+ base_url = f"http://{host_port}"
1340
1341
# Headers
1342
headers = None