Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,35 @@
|
|
| 1 |
---
|
| 2 |
license: creativeml-openrail-m
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: creativeml-openrail-m
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
**This model is trained for 5 characters from Saving 80,000 Gold in Another World for My Retirement (老後に備えて異世界で8万枚の金貨を貯めます)**
|
| 6 |
+
|
| 7 |
+
Most information along with examples can be found on the civitAI model page
|
| 8 |
+
https://civitai.com/models/17336/roukin8-character-lohaloconfullckpt-8
|
| 9 |
+
|
| 10 |
+
#### Usage
|
| 11 |
+
|
| 12 |
+
There are three formats: loha, locon, and full checkpoint
|
| 13 |
+
The first two are part of the [LyCORIS](https://github.com/KohakuBlueleaf/LyCORIS) project and can be used as lora in a1111-webui once this [extension](https://github.com/KohakuBlueleaf/a1111-sd-webui-locon) is installed
|
| 14 |
+
|
| 15 |
+
You probably want to separate names by ; as this is how the model is trained. E.g.
|
| 16 |
+
```
|
| 17 |
+
Sabine; YamanoMitsuha; multiple girls, 2girls, bow, sliding doors, apron, dress, indoors, holding, bowtie, smile, open door, door
|
| 18 |
+
```
|
| 19 |
+
|
| 20 |
+
As usual, two characters are fine, but more are difficult
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
#### Dataset Description
|
| 24 |
+
|
| 25 |
+
The dataset is prepared via the workflow https://github.com/cyber-meow/anime_screenshot_pipeline
|
| 26 |
+
|
| 27 |
+
It contains 2948 roukin anime screenshots and 19279 regularization images.
|
| 28 |
+
The anime screenshots can be downloaded from civitai.
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
#### Training
|
| 32 |
+
|
| 33 |
+
- Common: clip skip 1, resolution 512, batch size 8
|
| 34 |
+
- Full checkpoint: 2.5e-6 cosine scheduler, Adam8bit, conditional dropout 0.08
|
| 35 |
+
- Loha/LoCon: 2e-4 constant scheduler, AdamW
|