![]() |
Verifies that all given hyperparameters are consistent.
tfl.kronecker_factored_lattice_lib.verify_hyperparameters(
lattice_sizes=None, units=None, num_terms=None, input_shape=None,
monotonicities=None, output_min=None, output_max=None
)
This function does not inspect weights themselves. Only their shape. Use
assert_constraints()
to assert actual weights against constraints.
See tfl.layers.KroneckerFactoredLattice
class level comment for detailed
description of arguments.
Args | |
---|---|
lattice_sizes
|
Lattice size to check against. |
units
|
Units hyperparameter of KroneckerFactoredLattice layer.
|
num_terms
|
Number of independently trained submodels hyperparameter of
KroneckerFactoredLattice layer.
|
input_shape
|
Shape of layer input. Useful only if units and/or
monotonicities is set.
|
monotonicities
|
Monotonicities hyperparameter of KroneckerFactoredLattice
layer. Useful only if input_shape is set.
|
output_min
|
Minimum output of KroneckerFactoredLattice layer.
|
output_max
|
Maximum output of KroneckerFactoredLattice layer.
|
Raises | |
---|---|
ValueError
|
If lattice_sizes < 2. |
ValueError
|
If units < 1. |
ValueError
|
If num_terms < 1. |
ValueError
|
If len(monotonicities) does not match number of inputs. |