Jianyi Wang
Jianyi Wang
The reference image is taken as input for the trainable encoder. You may refer to readme to generate the reference image. 
My bad. I have updated the related code. Thanks for your issue.
Currently the code does not support multi-GPU test. You may refer to the training script to modify it to support multi-GPU test or you can run the same script multiple...
Hi. You may refer to [here](https://github.com/pkuliyi2015/sd-webui-stablesr/issues/54). I am not familiar with it. If you can further improve this repo please let me know. I will add the link on the...
You may check [here](https://github.com/IceClear/StableSR/blob/main/ldm/models/diffusion/ddpm.py#L2598C39-L2598C50)
I do not think there is an obvious difference between them.
Please refer to the paper.
Hi. Which platform did you use? It should bot be the GPU problem. But the platform is not stable actually, so error may happen. I will have a check later.
Hi. Our code currently may not support CPU.
Hi. This model is jointly used with a pre-trained model, i.e., codeformer for face restoration. You may refer to our latest paper for details. I may update the details later....