Commit
•
fd65d2b
1
Parent(s):
769aae4
Update README.md
Browse files
README.md
CHANGED
|
@@ -15,7 +15,15 @@ A dreambooth finetune of stable diffusion 1.5 model that will output stuff in an
|
|
| 15 |
# Usage
|
| 16 |
Keep the words `anime pencil concept style` towards the beginning of your prompt to invoke the finetuned style.
|
| 17 |
|
| 18 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
## Anime-Pencil-Diffusion-V3
|
| 21 |
Trained for 12000 steps, constant learning rate of 0.0000005 on 80 images with 1000 images of `illustration style` for regularization
|
|
|
|
| 15 |
# Usage
|
| 16 |
Keep the words `anime pencil concept style` towards the beginning of your prompt to invoke the finetuned style.
|
| 17 |
|
| 18 |
+
# V4 upload in progress
|
| 19 |
+
I'd say wait for v5 or use earlier versions, since I'm dumb and didn't train the text encoder nor used regularization images for v4 training
|
| 20 |
+
Wasted lots of electricity on v4, hope v5 is way better in terms of style
|
| 21 |
+
|
| 22 |
+
## Anime-Pencil-Diffusion-V4
|
| 23 |
+
Trained for 160000 steps, constant learning rate of 0.000001 on 526 images with 0 images for regularization and no text encoder training
|
| 24 |
+
|
| 25 |
+
### Examples generated by the v4 model
|
| 26 |
+
<img src="https://huggingface.co/yehiaserag/anime-pencil-deffusion/resolve/main/example-v4.jpg"/>
|
| 27 |
|
| 28 |
## Anime-Pencil-Diffusion-V3
|
| 29 |
Trained for 12000 steps, constant learning rate of 0.0000005 on 80 images with 1000 images of `illustration style` for regularization
|