lora-scripts
lora-scripts copied to clipboard
Akegarasu
SD-Trainer. LoRA & Dreambooth training scripts & GUI use kohya-ss's trainer, for diffusion model.
手动选择配置就是方便在几套配置间切换(主要是人物lora和画风lora还有概念lora之间配置不同) 训练队列就是希望能一次性排队多个训练,进行完一个就继续进行下一个训练,排队方式可以通过配置文件完成,也就是队列执行选定的多个配置文件 还有一个目前用着比较难受的点是现在不能在gui里面终止训练,想停止已经开始的训练只能关闭控制台然后重启.如果能直接在gui里面终止训练就好了
请教一下我把训练脚本train.sh中is_v2_model设成1，parameterization=1训练sd2.1-768，但是loss一直不收敛（>0.3），请教一下是还需要改哪里吗？
Some people (as me) are running all of those stable diffusion scripts on distant computers in order to have more performant devices at disposition. Since [AUTOMATIC1111/stable-diffusion-webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui) and [bmaltais/kohya_ss](https://github.com/bmaltais/kohya_ss) are using...
点击直接开始训练后报错： ``` Traceback (most recent call last): File "/data/home/duval1024/lora-scripts/./sd-scripts/train_network.py", line 17, in import library.train_util as train_util File "/data/home/duval1024/lora-scripts/sd-scripts/library/train_util.py", line 56, in import albumentations as albu File "/data/home/duval1024/lora-scripts/venv/lib/python3.10/site-packages/albumentations/__init__.py", line 5, in...
   显示一个大大的图标，没有界面
when train I get the following problem. Have anyone have same problem? steps: 0%| | 0/20880 [00:00
``` running training / 学習開始 num train images * repeats / 学習画像の数×繰り返し回数: 78 num reg images / 正則化画像の数: 0 num batches per epoch / 1epochのバッチ数: 26 num epochs / epoch数:...
