i downloaded a different version of my source model, one named fp16 (and it was a lot smaller)- this
i downloaded a different version of my source model, one named fp16 (and it was a lot smaller)- this one does not give errors anymore, I tried 2 epochs but when i apply the lora in automatic1111 but it doesn't seem to change the image. tensorboard did show the loss going down signficantly though, so i'm not really sure why the default installation (torch2, with latest code of kohya_ss) didn't seem to learn anything. I then found this part from your notes: https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/Generate-Studio-Quality-Realistic-Photos-By-Kohya-LoRA-Stable-Diffusion-Training-Full-Tutorial.md where it says "if it works but not learns" do this-- i checked out the git hash that you pasted there and its using torch1.12.1 instead of torch 1.13 like it says right above it-- are these still good instructions to use when the lora doesn't seem to be learning?
GitHub
Stable Diffusion, SDXL, LoRA Training, DreamBooth Training, Automatic1111 Web UI, DeepFake, Deep Fakes, TTS, Animation, Text To Video, Tutorials, Guides, Lectures, Courses, ComfyUI, Google Colab, R...