I delved back into the world of Stable Diffusion yesterday, this time to focus on training models. I did not have success. Just trying to figure out how to run Dreambooth was confounding. I tried running it locally as well as in Colab, but a nebulous web of commands and dependencies meant that my first couple attempts were a waste.
Running these training models takes several hours, so trial and error here is monumentally slow. And compatibility with the M1 Mac is a bit difficult as well. I almost wish I still had my six-GPU mining machine to run these models against. Oh, well I’ll have to try again next week. I spent way to much time working on this instead of doing what I was supposed to do yesterday, so I’ll need to see what I can do without tying myself up all day.