ValueError: Expected both input arguments to be normalized tensors with shape [N, 3, H, W]. Got input with shape torch.Size([10, 3, 192, 192]) and torch.Size([10, 3, 192, 192]) and values in range [tensor(nan, device='cuda:3', grad_fn=<MinBackward1>)

[Mrguanglei]

Hello developer, I have a problem now. When I got to 1000 steps, ValueError: Expected both input arguments to be normalized tensors with shape [N, 3, H, W]. Got input with shape torch.Size([10, 3, 192, 192]) and torch.Size([10, 3, 192, 192]) and values in range [tensor(nan, device='cuda:3', grad_fn=), tensor(nan, device='cuda:3', grad_fn=)] and [tensor(-1., device='cuda:3'), tensor(1., device='cuda:3')] when all values are expected to be in the [-1, 1] range. "Error. I froze the VIT and retrained, but this error still occurred. Then I remove the logs and redo the operation. This error occurred in step 2. How to solve this problem, I fell into thinking. Hope to get your help