LoRA — Universal SDXL LoRA
Generated by Universal SDXL LoRA Training Workflow v1.2.0.
Info
- Base model:
Wh1teSpaces/OrthogonalDiffusion-XL/OD_XL_v0_3_IllustriousStyle_Boost_fp16.safetensors - Trigger word:
none - Dataset repo:
Bl4ckSpaces/Ejen-Kim-Char-Tagged-v2 - Images:
16 - Resolution mode:
native_pad8 - Batch size:
2 - Effective epochs:
125 - Total steps:
1000 - LoRA rank/alpha:
8/16 - LR:
0.0001 - Optimizer:
PagedAdamW8bit - Pruning mode:
skip - Pruning rounds:
0 - Unique tags removed:
0 - Occurrences removed:
0 - Best loss:
0.042633@ epoch57 - Final loss (last 20):
0.110357
Pruning details
See tags_summary.md, tags_report.json, caption_audit.json.
Usage
Trigger:
(no trigger)
Model tree for Wh1teSpaces/Ejen-Kim
Base model
Wh1teSpaces/OrthogonalDiffusion-XL