Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- lora
|
| 4 |
+
- diffusers
|
| 5 |
+
- qwen-image
|
| 6 |
+
- 20b
|
| 7 |
+
- bf16
|
| 8 |
+
base_model: Qwen/Qwen-Image
|
| 9 |
+
license: cc-by-nc-4.0
|
| 10 |
+
---
|
| 11 |
+
A Qwen-image 20b LoRA for making BBWs (big beautiful women).
|
| 12 |
+
|
| 13 |
+
Note:
|
| 14 |
+
- Qwen seems to be sensitive to pulling the image off the LoRA style if you prompt for a style not in the dataset, i.e. "chibi". The LoRA was trained on photographs so it will do best with realisim prompts.
|
| 15 |
+
- steps 40, cfg 3.5 seemed to make good images
|
| 16 |
+
- I tested with LoRA strength 1.0
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
Trained using diffusion-pipe on 147 images to epoch 95 using the 'automagic' optimizer, rank 32.
|