If you have an NVIDIA GPU, ensure you have installed the appropriate CUDA toolkit and cuDNN versions compatible with your PyTorch installation. python main.py train \ --batch-size 16 \ ...