Trained with 45 image using 10 epoch
Adafactor + cosine
I think offset is slightly too much this time so some generation feels brighter
More Lora will be trained, you can also comment on characters you want, I will utilize both my gpu and civit ai on site training. The Sd3.5 medium training process is stuck because my internet is too slow and I have huge dataset still waiting to be uploaded.
If you like my work please consider supporting me: ko-fi.com/suzushi2024
Also building an alternative platform for model hosting and image sharing