GypsyBoostingMachine flexible gradient boosting framework Main features: user-defined loss function (via using AD) user-defined base learners overfitting control during boosting iterations via test\val split ordering support