Yinghao Zhang

I don't think this implementation consistent with the original paper (RAKI). In `Network.py`, the RAKI network are build by, ``` net = nn.Sequential( nn.Conv2d(in_channels=2 * self.channels_in, out_channels=128, kernel_size=[3, 5], padding=[1,...

linalg.svd - complex64 - NaN - bug

5

### Issue type Bug ### Have you reproduced the bug with TensorFlow Nightly? Yes ### Source source ### TensorFlow version tf nightly, 2.13, 2.9 using docker container ### Custom code...

questions about downsampling (blurring voxels)

1

### The following pic is a screenshot of your excellent paper.  ### Questions - what is the exact cut-off frequency of the HR signal and why this frequency has...

As in the code, the arg `learnable_input_prompt` is set to be `False` in default, and when i check the whole model `PromptMR`, i find that there is no `learnable_input_prompt` arg,...

现在 CZSC 类里面只有分型和笔，画图出来也没有中枢，是不是在这个类里面加上类似 FX_ZS(分型中枢) 和 BI_ZS(笔中枢) 更好一点，这样出图以及相应用到中枢的信号的编写也更方便直接一点？ 我自己大概实现了一下，如果项目需要的话，我可以提个PR合并一个新的 CZSCwZS (CZSC with ZS) 的类进去

In the code, β times `zq-z.detach()`, https://github.com/sczhou/CodeFormer/blob/e878192ee253cfcc8f19e29d3307c181501f53ae/basicsr/archs/vqgan_arch.py#L55 but i think it should times the commitment loss `zq.detach()-z` to control the learning of Encoder, referred to the CoderFormer and VQVAE papers....