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 10, 09:00 UTC
[Core] Use weights_only=True with torch.load (#32045)
Signed-off-by: Russell Bryant <rbryant@redhat.com>
Russell Bryant committed
Jan 9, 2026 at 19:28 UTC
1963245ed15a9f232ac3b763c7a03351d77ae799
1 file changed
+1
-1
vllm/model_executor/model_loader/tensorizer.py
+1
-1
@@ -762,7 +762,7 @@ def tensorize_lora_adapter(lora_path: str, tensorizer_config: TensorizerConfig):
762
if tensor_path.endswith(".safetensors"):
763
tensors = safetensors.torch.load_file(tensor_path)
764
elif tensor_path.endswith(".bin"):
765
- tensors = torch.load(tensor_path)
765
+ tensors = torch.load(tensor_path, weights_only=True)
766
else:
767
raise ValueError("Unsupported file: %s", tensor_path)
768