Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,48 @@
|
|
1 |
---
|
2 |
license: creativeml-openrail-m
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: creativeml-openrail-m
|
3 |
---
|
4 |
+
|
5 |
+
---
|
6 |
+
license: creativeml-openrail-m
|
7 |
+
---
|
8 |
+
|
9 |
+
This model is dreamboothed on four concepts from yamanosusume
|
10 |
+
|
11 |
+
**Prompts:**
|
12 |
+
|
13 |
+
1. aohina yuri kurauehinata yukimuraaoi 2girls
|
14 |
+
2. yukimuraaoi girl
|
15 |
+
3. kurauehinata girl
|
16 |
+
4. aobakokona girl
|
17 |
+
|
18 |
+
---
|
19 |
+
|
20 |
+
**Training details:**
|
21 |
+
|
22 |
+
- Trained with [TheLastBen's fast-DreamBooth notebook](https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast-DreamBooth.ipynb)
|
23 |
+
- data set: 198 concept images, I cannot remember whether I used reg images or not (it seems not)
|
24 |
+
- learning rate 1e-6 for 15500 steps
|
25 |
+
- text encoder rate 10%
|
26 |
+
|
27 |
+
---
|
28 |
+
|
29 |
+
**Problems:**
|
30 |
+
|
31 |
+
- As always, it is hard to get interactions between multiple characters even though I explicitly trained for it.
|
32 |
+
- For some mysterious reasono, when generating images for aohina it can hardly get the eye colors right (maybe should train at clip_skip=2?).
|
33 |
+
|
34 |
+
--
|
35 |
+
|
36 |
+
**Example generations (with inpainting):**
|
37 |
+
|
38 |
+

|
39 |
+

|
40 |
+

|
41 |
+

|
42 |
+

|
43 |
+

|
44 |
+

|
45 |
+

|
46 |
+

|
47 |
+

|
48 |
+
|