Update README.md
Browse files
README.md
CHANGED
@@ -33,6 +33,37 @@ An anime diffusion model finetuned on Würstchen V3.
|
|
33 |
<img class="image" src="https://cdn-uploads.huggingface.co/production/uploads/6456af6195082f722d178522/uua4L9aaqJ0LI8gYv4xmC.png" width="320">
|
34 |
</table>
|
35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
|
37 |
## WD Tags:
|
38 |
|
|
|
33 |
<img class="image" src="https://cdn-uploads.huggingface.co/production/uploads/6456af6195082f722d178522/uua4L9aaqJ0LI8gYv4xmC.png" width="320">
|
34 |
</table>
|
35 |
|
36 |
+
## Training:
|
37 |
+
|
38 |
+
**GPU used**: 7x Nvidia H100 80GB SXM5
|
39 |
+
### Stage C
|
40 |
+
| parameter | value |
|
41 |
+
|---|---|
|
42 |
+
| **amp** | no |
|
43 |
+
| **weights** | fp32 |
|
44 |
+
| **save weights** | fp32 |
|
45 |
+
| **resolution** | 1024x1024 |
|
46 |
+
| **effective batch size** | 84 |
|
47 |
+
| **unet learning rate** | 2e-6 |
|
48 |
+
| **te learning rate** | 1e-7 |
|
49 |
+
| **optimizer** | AdamW 8bit |
|
50 |
+
| **images** | 6M * 2 captions per image |
|
51 |
+
| **epochs** | 1 |
|
52 |
+
|
53 |
+
**GPU used**: 1x Nvidia H100 80GB SXM5
|
54 |
+
### Stage B
|
55 |
+
| parameter | value |
|
56 |
+
|---|---|
|
57 |
+
| **amp** | no |
|
58 |
+
| **weights** | fp32 |
|
59 |
+
| **save weights** | fp32 |
|
60 |
+
| **resolution** | 1024x1024 |
|
61 |
+
| **effective batch size** | 8 |
|
62 |
+
| **unet learning rate** | 8e-6 |
|
63 |
+
| **te learning rate** | none |
|
64 |
+
| **optimizer** | AdamW |
|
65 |
+
| **images** | 120K |
|
66 |
+
| **epochs** | 6 |
|
67 |
|
68 |
## WD Tags:
|
69 |
|