params = {
    'n_estimators' : 10000,
    'objective' : 'binary',
    'learning_rate' : 0.05,
    'reg_alpha' : 0.1,
    'reg_lambda' : 0.1,
    'seed' : 50,
    'metric' : 'auc',
    'verbose' : 100
}

for i in range(2, 12):
    train_bin = lgb.Dataset('train_bin_{}.bin'.format(i))
    valid_bin = lgb.Dataset('valid_bin_{}.bin'.format(i))

    model = lgb.train(params, train_bin, valid_sets=[valid_bin], early_stopping_rounds=100)
    print(model.best_iteration_)