torch.cuda.DeferredCudaCallError: CUDA call failed lazily at initialization 报错

完整的报错信息为：

torch.cuda.DeferredCudaCallError: CUDA call failed lazily at initialization with error: device >= 0 && device < num_g
pus INTERNAL ASSERT FAILED at "../aten/src/ATen/cuda/CUDAContext.cpp":50, please report a bug to PyTorch.

后来发现是‘CUDA_VISIBLE_DEVICES‘和代码中冲突了，其中在代码中包含了：

import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1,2,3'

和环境变量冲突了，将CUDA_VISIBLE_DEVICES的环境变量注释掉解决了问题

unset CUDA_VISIBLE_DEVICES