Okay so after getting helpful advice on creating a model Lora I manage to get it right! So here is a third version of Nita. Better and more accurate!
894234_training_data.zip
Nita_Bear_V3.safetensors