After several attempts, I finally managed to create this LORA, all thanks to the style of the artist Hunternif.
God bless him.
I personally generate using:
40 steps
dpmpp_2m
sgm_uniform.
My favorite model: plantMilkModelSuite walnut
In this version, I used only 16 images from the artist's entire gallery. I managed to achieve the model's flexibility thanks to one piece of advice that completely changed my vision of creating models. This model was created with 120 epochs specified, and this is epoch 103, by the way. To my surprise, the final epoch turned out to be heavily overfitted, but this one is perfect. At least, that's how it is in my hands. Others might have different feelings. In any case, I wish everyone all the very best.
Files
1807280_training_data.zip
Hunternif16-000103.safetensors