My First ZIT LoRA - Learning Experience

This is my first attempt at training a ZIT LoRA model, and I'm sharing it as a learning experience for the community.

Training Details:

- Dataset: 500+ images
- Training tool: AI Toolkit with helper LoRA
- Steps: 15,000
- Training time: 20+ hours on RTX 5060 Ti
- Settings: Default configuration

Current Issues:

- Inconsistent tag recognition
- Random generation results
- May require specific prompting techniques

Looking for Feedback:

I'm still learning the training process and would greatly appreciate any advice from experienced trainers. If you have suggestions on improving tag learning, optimal step counts, or dataset preparation, please share your insights in the comments!

Note: This model may not perform as expected. Consider it an experimental release while I refine my training approach.

If you've read this far, here's the full list of tags I attempted to train (most don't work):

tiny breasts, small breasts, medium breasts, big breasts, large breasts, huge breasts, round breasts, teardrop breasts, perky breasts, mild sagging breasts, sagging breasts, very sagging breasts, uneven breasts, firm breasts, soft breasts, natural breasts, fake breasts, breasts apart, tan

small areolas, medium-small areolas, medium areolas, big areolas, large areolas, very large areolas, round areolas, oval areolas, pigmented areolas, dark areolas, light areolas, brown areolas, pink areolas, skin color areolas, smooth areolas, bumpy areolas, raised areolas, flat areolas

small nipples, medium-small nipples, medium nipples, big nipples, large nipples, very large nipples, short nipples, long nipples, pointed nipples, round nipples, flat nipples, inverted nipples, dark nipples, light nipples, brown nipples, pink nipples, dark pink nipples, pierced nipples, smooth nipples, bumpy nipples,