guomc
Results
1
comments of
guomc
It is because you enable 'torch.utils.checkpoint' to save GPU memory. If you want to update learnable modules in checkpoint function, you must ensure your input tensor.requires_grad = True since Tune-A-Video...