## vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine

- 種別: advisory
- 重大度: medium
- 関連技術: python
- CVE: CVE-2026-115913
- 公開日時: 2026-10-28T21:15:41.000Z
- 情報源: GitHub Advisory Database (https://github.com/advisories/GHSA-8397-8qf6-f2gm)
- パーマリンク: https://vulnpulse.app/items/ghsa-ghsa-8397-8qf6-f2gm

影響: pip:vllm < 0.32.0 → 0.32.0
The Phi-4 multimodal audio encoder (phi4mm_audio.py / phi4mm_utils.py) runs its conformer attention in eager mode by default: the official HF config of microsoft/Phi-4-multimodal-instruct does not set use_pt_scaled_dot_product_attention, so the default False applies (phi4mm_audio.py:166, :876; phi4mm_utils.py:1669). The eager path fully materializes the attention probability matrix of shape [n_chunks, 16, 500, 500] per layer — and then stores it on the module: self.attn = attn (phi4mm_utils.py:…