I used Realistic Vision V5.1 to train the LORA. From my perspective, Realistic Vision 5.0 is better at generating realistic outputs.