| very interesting that the style fails to transfer to outside of school environment where most of the sample images are placed. Right most of grid is a 1 dim lora which generalizes to out of dataset. | |
| Would be interesting to train with higher dim or more aggressive settings |