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, 11:00 UTC
Don't default construct `ModelConfig` when default constructing `VllmConfig` (#17943)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Harry Mellor committed
May 10, 2025 at 14:23 UTC
68311891f5036a4faac89ead7dd40826d18da0b1
1 file changed
+3
-1
vllm/config.py
+3
-1
@@ -3954,7 +3954,9 @@ class VllmConfig:
3954
simplifies passing around the distinct configurations in the codebase.
3955
"""
3956
3957
- model_config: ModelConfig = field(default_factory=ModelConfig)
3957
+ # TODO: use default_factory once default constructing ModelConfig doesn't
3958
+ # try to download a model
3959
+ model_config: ModelConfig = None # type: ignore
3960
"""Model configuration."""
3961
cache_config: CacheConfig = field(default_factory=CacheConfig)
3962
"""Cache configuration."""