Update grimoire/release_notes/v1.0.md
Browse files
grimoire/release_notes/v1.0.md
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
|
| 5 |
A LoRA based on 4267 [Grimoire A: Private Grimoire Magic Academy](https://grimoire.applibot.co.jp/) cards, recognizing 60+ characters to varying degrees.
|
| 6 |
|
| 7 |
-
A request by Civit user themonky007. The model is far from complete but there'
|
| 8 |
|
| 9 |
Lesson of the day: suitable learning rates can differ significantly between datasets (this uses 2e-4 while the Idolmaster models use 1e-3).
|
| 10 |
|
|
|
|
| 4 |
|
| 5 |
A LoRA based on 4267 [Grimoire A: Private Grimoire Magic Academy](https://grimoire.applibot.co.jp/) cards, recognizing 60+ characters to varying degrees.
|
| 6 |
|
| 7 |
+
A request by Civit user themonky007. The model is far from complete but there're currently no plans to further improve it. [The dataset](https://huggingface.co/datasets/gustproof/sd-data/blob/main/grimoire_cards_notext_denoised.zip) is available for those willing to take on the task.
|
| 8 |
|
| 9 |
Lesson of the day: suitable learning rates can differ significantly between datasets (this uses 2e-4 while the Idolmaster models use 1e-3).
|
| 10 |
|