tanbuzheng


How to achieve actual acceleration of dw conv？

1

Hi, can you tell me how to implement actual acceleration of dw conv？

您好，在实现代码的过程中，我发现dw conv比较耗时，怎么实现它的实际加速呢？

Can you provide pre-trained models about image inpainting task

3

Thank you for creating such an outstanding work. We hope to compare with this work in our future work. So can you provide the pretrained models of image inpainting on...

Dear author, Thanks for sharing the code. I am greatly interested in your work. In my latest work, I hope to borrow ideas from your work. I have a request...

Question about training speed

4

Hi，I tried to use HiLo block on the image restoration task, but I found that the training speed of HiLo is much slower compared to Swin. Specifically, we mainly adopt...

Question about the defintion of Pruning loss

1

loss = torch.sum(torch.abs(noisy_img - img_tensor)) I hope to know why define pruning loss like above and the reason for minimizing this loss.

作者您好！我看了您的文章，觉得写的很好，但有点疑问。 关于线性注意力的实现感觉和performer有相似之处，但不知道为什么要这么改进？希望得到您的解答。

Dear author, Thanks for sharing the code. I am greatly interested in your work. I have a question for you and would like your reply. In P-VQVAE, the masked and...

A question about vocab_size in token embbeding

23

Hi！ When I read your source code, I found you set vocab_size = self.codebook_size + 1000 + 1 in token embbeding stage. Why not directly set vocal_size=self.codebook_size? What does the...