ai-toolkit
ai-toolkit copied to clipboard
How to train a feature in a clothe
Hi,
I'm trying to train a feature in a clothe (a crease) I've managed to find 50 images of it.. I was training for the last couple of days but always failed to achieve tge results.
- Sample images always look the same even over 4k steps
- Sometimes my training images showed panties while there was no implications in my caption or my sample prompt for panties
- Ive changed the settings from the default 24gb yaml file, lr, steps, resolution, but still nothing worked.
- I tried with no vaption, didn't work.
- I tried a caption just with the trigger word, didn't work
- I tried no trigger word, didn't work
- Tried to imply for the crease in my caption, didn't work
I'm sure I'm doing something wrong..
Can you guide me a bit?
Thanks.
Try doing inference in ComfyUI with your LoRAs. For me, the samples in the AI-toolkit don’t seem to work, but when I use the same LoRA in ComfyUI, the resulting images look perfect.