Update README.md
Browse files
README.md
CHANGED
|
@@ -39,7 +39,14 @@ license: apache-2.0
|
|
| 39 |
|
| 40 |
A face LORA of the famous https://www.twitch.tv/thinknomore.
|
| 41 |
|
| 42 |
-
The LORA is trained using Ralph Fiennes as an instance prompt, using [Kohya v22.2.2](https://github.com/bmaltais/kohya_ss/releases/tag/v22.2.2)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
|
| 44 |
More training details in the [checkpoint](https://huggingface.co/hemular/thinknomore-lora/tree/main/checkpoints) directory. The chosen "best" checkpoint is #7
|
| 45 |
|
|
|
|
| 39 |
|
| 40 |
A face LORA of the famous https://www.twitch.tv/thinknomore.
|
| 41 |
|
| 42 |
+
The LORA is trained using Ralph Fiennes as an instance prompt, using [Kohya v22.2.2](https://github.com/bmaltais/kohya_ss/releases/tag/v22.2.2)
|
| 43 |
+
|
| 44 |
+
Key training attributes:
|
| 45 |
+
|
| 46 |
+
- non-null starting prompt (Ralph Fiennes)
|
| 47 |
+
- `0.00001` learning speed
|
| 48 |
+
- 23 high-res non-cropped images.
|
| 49 |
+
- `1024_MAN` regularization images of the great [aitrepreneur](https://www.patreon.com/aitrepreneur) ([guide](https://www.youtube.com/watch?v=N_zhQSx2Q3c))
|
| 50 |
|
| 51 |
More training details in the [checkpoint](https://huggingface.co/hemular/thinknomore-lora/tree/main/checkpoints) directory. The chosen "best" checkpoint is #7
|
| 52 |
|