# Fizgig MOSS lora training help — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/a1e17a940c13f201

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-20 02:36 ingested 2026-10-21 02:38

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xagikk/fizgig_moss_lora_training_help/)

## retained text



```
Has anyone had good success with Fizgig MOSS image lora training? I've tried 30-70 epoch, .25-.5 MP, 20imgs-40imgs, high quality face only or mix of high quality body and face, preset fast, alpha rank 16, learning rate 0.0001, and block 22 included in the workflow in comfyui. I get ok likeness occasionally with 30 step workflow using no loras but the character lora and very meh likeness using 8 step lora workflow using 12 steps in my sampler and the character lora. 

I know it just came out so people are still fine tuning, but I have checked multiple threads on reddit and tried changing multiple settings. Sadly nothing seems to be working. I would like to be able to use the 8 step lora workflow along with my character lora since I'm using an rtx3080 and it takes awhile to process high MP 30 step images. 

Any help appreciated
```



document fetched 2026-10-21 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xagikk/fizgig_moss_lora_training_help
