This is the first version of this LoRA for Sayuri Mattar. It was trained with only 1100 steps over 38 mixed body/face pictures. I may create a better version in the future, but for now, I am only testing and learning how to create LoRAs.
Description
Test version
FAQ
Comments (2)
Could you share the training settings you used?
Most of the settings I used can be found in the post description. 38 mixed body/face auto-tagged images, 50 epochs, 1100 steps, and clip skip 2. It was trained with Pony Realism v2.1 Main + VAE. The only other parameter you cannot find in the post is the network rank, and I don't remember what value I used. But since it was the first test LoRA I made for Sayuri, I probably used a low value, between 8 and 32. So far, I have failed to get good results with higher values for network rank.