I have seen a few people on the Stable Diffusion subreddit who are having difficulty getting the new 2.1 model setup. Here is a quick guide to get the new model setup on AUTOMATIC1111 (it should work for most other web ui’s as well).
Before you begin, make sure AUTOMATIC1111 is updated to the latest build. You can do this by opening the directory where you installed it, opening CMD and run:
git pull
Now you’ve got the latest version, download the v2.1 checkpoint file from HuggingFace if you haven’t already.
https://huggingface.co/stabilityai/stable-diffusion-2-1/tree/main
Download the “v2-1_768-nonema-pruned.ckpt” version and place it in your Stable Diffusion models folder. Unless you intend to train your own models in which case you should download the ema version.
The next file you need is the .yaml configuration. The easiest way to download this is right clicking on the GitHub link and Save As, make sure you save it as a .yaml file and not .txt.
Rename the .yaml file to “v2-1_768-nonema-pruned.yaml” and move it into your models folder where you placed the .ckpt file earlier.
Now you should be able to launch Stable Diffusion with AUTOMATIC1111 and select the new v2.1 model as your checkpoint. If you have a weaker graphics card and are having difficulty generating images you may need to edit your webui-user.bat file with the following.
Replace:
set COMMANDLINE_ARGS=With:
set COMMANDLINE_ARGS=--precision full --no-half --medvramThis should free up some memory and stop Stable Diffusion from being so resource intensive, it may take longer to generate images however. It seems that the 2.0 and 2.1 models are harder to run compared to 1.4 and 1.5.
