This is a test model.
I recommend a LoRA strength of about 0.7.
The model will be re-trained after reviewing the training data.