I've wanted to create this Lora for a long time out of curiosity about how learning can be done mainly on crystals, and here I just came across a few artists whose work was beautiful in itself, but I recognize this version as valid. The fact is that I made an elementary mistake: I did not remove the signatures of the authors, and there are a couple of not the best crystals in the dataset, which negatively affected the result. Nevertheless, these results for the first test time are satisfactory for my personal purposes.