Trained on 264 ethically-sourced images of "kintsugi" objects, at 640x640 resolution, using the Prodigy optimizer, batch size 3, learning rate 1.05 (Cosine), for ten epochs; the output LoRA from the 8th epoch of training is presented here. Dim/Alpha=32/16
Trained on 264 ethically-sourced images of "kintsugi" objects, at 640x640 resolution, using the Prodigy optimizer, batch size 3, learning rate 1.05 (Cosine), for ten epochs; the output LoRA from the 8th epoch of training is presented here. Dim/Alpha=32/16