Verified: 2 years ago
PickleTensor
Type | |
Stats | 2,074 |
Reviews | (309) |
Published | Apr 3, 2023 |
Base Model | |
Trigger Words | dacishuwang |
Hash | AutoV2 233500D116 |
Model
同时使用ti和lora你能在绝大部分模型中都得到不错的效果。
Using ti and lora together you can get good results in most models.
Why ti+lora?
ti和lora一起用能在照片基础的模型上工作的更好,而且不用写一大堆tag。
有时候你不能加太多人物lora的权重,就需要人物ti来补充人物特征。
Use both TI and Lora works very well with photrealistic models,and you don't neet to write a lot of tags.
Sometimes you can't add too much weight to the character lora, so you need character ti to complement the character characteristics.
lora+tags:
ti+lora:
Use lora block weight to limit lora
Problems:
树王的图好少!由于图不足,当初训练的时候ti只给了12个向量,现在会有不小的概率生成出四只耳朵或者两根辫子,或者叶子头饰的位置不对。
Rukkhadevata's picture is so few! Due to a lack of graphics,I trained ti I only use 12 vectors.Sometimes well give four ears or two braids, or a leaf ornaments in the wrong place.