This lora is about background. The city is china city. Especially, the trained images were from Qinzhou, China. That place was very good. People welcomed me. I have good memories...
Caution
Sometimes, the lora will render no humans image although you input prompt about human.
Sometimes, the image render rotated background image. It will be fixed if you draw a human. I guess that the Stable Diffusion cannot recognize what image is the correctly rotated if there is no human.
Description
FAQ
Comments (6)
中国永远欢迎你，我的兄弟！
China will always welcome you, my brother！
Thanks Bro!!
Can you give me a detail of the training setting
Um... I want to respond your question, but I need more detail in your question. About how many photos input to training? or number of epochs.?.. etc.
I input train image that I took pictures in China, when I had been Qinzhou, China for business trip. The number of photo I used is 100. Number of Repeat : 3. Epochs : 15. Batch size : 3.
Thank you very much, I still have one more question, do you use the v1-5-pruned-emaonly.safetensors for training?
@326922943737 no, i didn't... I don't know what it is... sorry