解决RuntimeError: CUDA out of memory，reserved memory is ＞＞ allocated memory

解决RuntimeError: CUDA out of memory，reserved memory is >> allocated memory

报错信息： torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.59 GiB (GPU 0; 23.68 GiB total capacity; 3.86 GiB already allocated; 769.69 MiB free; 5.96 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

解决方法：
这里要注意是你的reserved memory is >> allocated memory可以使用的，但也不是一定有用，最后最后如果还不行的话可以试试（参数128可以调整）

代码如下：

import os
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "max_split_size_mb:128"

参考博客

https://zhuanlan.zhihu.com/p/581606031