supervised_training_hyperparams.ini

[hyperparameters]
    learning_rate = 5e-3
    milestones = [5000, 21000, 32000]
    gamma = 0.5
    momentum = 0.9
    c = 1e-4
    epochs = 1
    batch_size = 256

[imitation_constraints]
    min_white_elo = 2300
    min_black_elo = 2300
    worse_games = 25