Add user friendly GUI for training Dreambooth by cpu?

[NaughtDZ]

Compared with the graphics card with 16G or even 32G video memory, I think it is an easy and low-cost way to obtain 32G memory. Through the CPU training model, it is only a few hours or dozens of hours, which is acceptable for me personally. And obviously, the specific images obtained through Dreambooth must be more accurate.
This project generation can use CPU to train Dreambooth. However, I failed to run this project locally. It is not because of insufficient memory, but because of the errors in various script files and various environment configurations, I was deterred.

https://github.com/andreae293/Dreambooth-Stable-Diffusion-cpu

或许您可以在网页中构建一个用户友好的图像化训练界面？
Perhaps you can build a user-friendly graphical training interface in the webpage?

[ClashSAN]

edit: I found it - #2002 (comment)
I once saw people saying they had gotten dreambooth running on cpu in the old dreambooth PR. You may want to download his version from a specific point in time to try. I think he is using the .ckpt architecture not the diffusers

[NaughtDZ]

https://github.com/d8ahazard/stable-diffusion-webui/tree/Dreambooth Training on CPU in webui works for me. download, install the dependencies and in the settings set to use the CPU

Thank you,I will try it.

[NaughtDZ]

https://github.com/d8ahazard/stable-diffusion-webui/tree/Dreambooth Training on CPU in webui works for me. download, install the dependencies and in the settings set to use the CPU

Sorry but no UI entry for training Dreambooth was found in the branch version you sent...Where can i get it?

[ClashSAN]

@NaughtDZ
see if you can get the commit around that time: #2002 (comment)

[NaughtDZ]

@NaughtDZ see if you can get the commit around that time: #2002 (comment)

Aha, at first I just used "git clone" to get the branch, but when I ran it, there was no Dreambooth option in the tab. When I chose to download it as a zip and overwrite it directly from the web page, it appeared. Thanks!

[vmajor]

https://github.com/d8ahazard/stable-diffusion-webui/tree/Dreambooth Training on CPU in webui works for me. download, install the dependencies and in the settings set to use the CPU

How did you do this?

Using --use-cpu all at runtime gives this when trying to run any model:

raise ValueError('Expected a cuda device, but got: {}'.format(device))
ValueError: Expected a cuda device, but got: cpu