I learned that training a model isn't about using as many images as possible, so I selected some images and retrained the model. (Use this method happily.)
artistStyleSamV3_v1.safetensors