...born March 21, 1972, a Polish film, television, voice and theatre actor, now available as a LoRa model 👀 █▬█ █ ▀█▀A 2.1 version of the Piotr Adamczyk LoRa - this one is more robust as I've got insight into many different parameters for training.A 1.0 version is performing way worse from my experience but if you really want - you can try it with the settings provided in "about this version".5 epochs on 22 photos, 20 repeats, based on <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0">stable-diffusion-xl-base-1.0</a> 2200 steps in total.Works best with ~1.3 strength_model and 1.1 strength_clip.For the best results, create an image at 1024x1024.For sampling I use 40 steps, DPMPP 3M SDE, exponential, with 3.0 cfg.Feel free to message me in case on any issues/suggestions.

Piotr Adamczyk

15 epochs on 22 photos, 7 repeats, based on stable-diffusion-xl-base-1.02310 steps in total.Works best with ~0.9-1.1 strength and portrait/closeup type images.For the best results, first create an image at maximum of 512x512 resolution and then upscale it later, larger resolutions seem to give worse results.For sampling I use 40 steps, DPMPP 3M SDE, exponential, with 5.0 cfg

Files with SHA256: 4557a44884de95b747f8880d47e55fddc1380c2966de6402ff9104ac9c874689