>>935892127
lowering from 10 to 1 = the model gets 1/10th of exposure to each image. meaning it wont be able to understand them enough to lock in what youre trying to train (under training)