setu.NLESearchSpace

class setu.NLESearchSpace(flow_type=('nsf', 'maf'), n_layers=(2, 12), nn_width=(32, 64, 128, 256), nn_depth=(1, 3), activation=('relu', 'tanh', 'silu'), learning_rate=(0.0001, 0.005), knots=(4, 16), interval=(2.0, 6.0))

Bases: object

Search space for NLE hyperparameter tuning.

Convention: tuple[int, int] or tuple[float, float] fields are sampled as uniform ranges (low, high). Untyped tuple fields are sampled as categorical choices.

Parameters:

• flow_type (tuple)

• n_layers (tuple[int, int])

• nn_width (tuple)

• nn_depth (tuple[int, int])

• activation (tuple)

• learning_rate (tuple[float, float])

• knots (tuple[int, int])

• interval (tuple[float, float])