CVE-2025-24357Published Jan 27, 2025 · Jan 27Updated Oct 28
vLLM allows a malicious model RCE by torch.load in hf_model_weights_iterator
vllm-project vllm
High7.5Fix available
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| vllm PyPI |
|
|
Original description
vLLM is a library for LLM inference and serving. vllm/model_executor/weight_utils.py implements hf_model_weights_iterator to load the model checkpoint, which is downloaded from huggingface. It uses the torch.load function and the weights_only parameter defaults to False. When torch.load loads malicious pickle data, it will execute arbitrary code during unpickling. This vulnerability is fixed in v0.7.0.
From the CVE record · GitHub_M