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 14, 05:30 UTC
[Bugfix] when set offline model running error (#23711)
Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
rongfu.leng committed
Aug 28, 2025 at 15:27 UTC
daa1273b14da5bdf643aa4b1bcbef3985b1edd75
1 file changed
+3
-1
vllm/entrypoints/utils.py
+3
-1
@@ -313,12 +313,14 @@ def log_non_default_args(args: Union[argparse.Namespace, EngineArgs]):
313
314
# Handle EngineArgs instance
315
elif isinstance(args, EngineArgs):
316
- default_args = EngineArgs() # Create default instance
316
+ default_args = EngineArgs(model=args.model) # Create default instance
317
for field in dataclasses.fields(args):
318
current_val = getattr(args, field.name)
319
default_val = getattr(default_args, field.name)
320
if current_val != default_val:
321
non_default_args[field.name] = current_val
322
+ if default_args.model != EngineArgs.model:
323
+ non_default_args["model"] = default_args.model
324
else:
325
raise TypeError("Unsupported argument type. " \
326
"Must be argparse.Namespace or EngineArgs instance.")