Torch models trained in higher versions report errors in lower versions

pttempted to read a PyTorch file with version 3, but the maximum supported version for reading is 2.

solve ：

First use the higher version torch Save again after reading .

model = torch.load('model_name.pt')

torch.save(model, 'save_path', _use_new_zipfile_serialization=False)

Then read in the lower version .

torch.load('model path', map_location='cpu')