During the training phase, your paper does not appear to calculate the loss of the image gradient, but this is in the code; Although there are mse_loss and simm_loss of calculating images in the code, they do not participate in backpropagation; In addition, I did not see the lite_transformer mentioned in the paper in the code. I am very confused by this, and I appreciate it if you can correct and answer these questions.
During the training phase, your paper does not appear to calculate the loss of the image gradient, but this is in the code; Although there are mse_loss and simm_loss of calculating images in the code, they do not participate in backpropagation; In addition, I did not see the lite_transformer mentioned in the paper in the code. I am very confused by this, and I appreciate it if you can correct and answer these questions.