
Have you ever wondered for how long you can finetune a model to learn a character? i did(it's a lot). This model is a extracted lora of PonyDiffusionV6XL finetuned with 124 images of gigi murin and tags from danbooru.
Trained for 200 epochs = 24800 steps
turned out surprisingly good, it can even draw her tail accurately.
it took a lot of time -_- (22 hours on a rtx 3060)
The captions used: gigimurindataset - Pastebin.com
描述:
Rank 32 lora extract
训练词语: 1girl,gigi murin,hololive,hololive english,pink eyes,brown hair,hoodie,bike shorts,ahoge
名称: GigiMurinPony32.safetensors
大小 (KB): 249018
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
原文链接:https://1111down.com/1071812.html,转载请注明出处