【QWEN-VL-2.5版本】本地部署，视频理解任务时报错：out of memory

各位专家好，
本地部署了Qwen2.5-VL-7B-Instruct，做视频理解，上传了一个34MB大小的视频。
问题是：这个视频显示了一个交通路口，画面存在抖动"\nAssistant:"

   本地部署报错信息如下：torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 40.80 GiB. GPU 0 has a total capacity of 79.15 GiB of which 15.61 GiB is free. Including non-PyTorch memory, this process has 63.54 GiB memory in use. Of the allocated memory 23.85 GiB is allocated by PyTorch, and 39.20 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation.  See documentation for Memory Management  (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

我本机的部署是：
双A100机器，显存80GB*2

按照报错信息，我阐释了设置PYTORCH_CUDA_ALLOC_CONF=expandable_segments
结果直接起模型直接core dump。