Lincoln Stein
Lincoln Stein
> Overall, makes sense to me. A few comments: > > * I smoke tested the lineart processor. I trust that you have tested them all? Yes, all tested successfully....
> Is the current watermarking mode what we want to commit to long term? Seems like it makes a notable impact on images; although, I’m not sure there’s a better...
The issue with `v_prediction` vs `epsilon` has been worked-around. The issue with unconvertible sd 1.5 controlnets has been worked-around. Loading of SDXL controlnet single files still does not work.
> Are you still following this PR @lstein ? I can take it over and get it across the line if you'd like I’m very happy to see renewed activity...
> Are you still following this PR @lstein ? I can take it over and get it across the line if you'd like @brandonrising I've removed the absolute path name...
Do you happen to have `convert_cache_dir` set in your `invokeai.yaml` file? Here's a possible fix for your specific problem: 1. Open the model manager tab 2. Select the model that...
I started working on this PR over the weekend. This is an impressive and highly functional piece of work that will make many Invoke users very happy. So far I've...
Also I can confirm that both in this PR and in `main`, when rendering a 1024x1024 image with an SDXL model, slice sizes, of 3, 6, 7 and 8 produce...
I had a quick look at this, but couldn't figure out the UI for adjusting the prompt weightings. I guess the prompt weight code isn't wired up to the UI...
@joshistoast Just checking in to see if you're still working on this?