Can multi-GPU training be used to attack the dataset?/是否可以使用多GPU对数据集进行攻击

[cgdrm]

Hello author, thank you for your wonderful work and contribution. When I attack graph data, I often encounter the problem of insufficient GPU memory, such as the Pubmed dataset. I would like to ask if it is possible to use multiple GPUs for the attack, and if so, how should I operate?
您好作者，感谢您的精彩工作和贡献。我在攻击图数据的时候经常遇到GPU内存不足的问题，例如Pubmed数据集。我想问是否可以使用多个GPU进行攻击，可以的话应该怎么操作？

RuntimeError: CUDA out of memory. Tried to allocate 24.00 MiB (GPU 0; 15.89 GiB total capacity; 14.17 GiB already allocated; 23.12 MiB free; 15.21 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF