def __init__(self, flow, latent_prior, verbosity=None):
self.flow = flow
self.latent_prior = latent_prior
self.loss = abstract_attribute()
self.logger = logging.getLogger(__name__).getChild(
self.__class__.__name__ + ":" + hex(id(self)))
self.set_verbosity(verbosity)
def __init__(self, *, d, transform, mask):
"""
Parameters
----------
d: int
dimension of the space
transform: function or :py:class:`InvertibleTransform <zunis.models.flows.coupling_cells.transforms.InvertibleTransform>`
bijective variable transform that maps the unit hypercube (in `d-sum(mask)` dimensions) to itself
given some parameters. It should have the same signature as
:py:meth:`InvertibleTransform.forward <zunis.models.flows.coupling_cells.transforms.InvertibleTransform.forward>`
mask: list of bool
indicates which variables are passed (y_N) through and which are transformed (y_M)
"""
super(GeneralCouplingCell, self).__init__(d=d)
self.mask = mask + [False]
self.mask_complement = not_list(mask) + [False]
self.transform = transform
self.T = abstract_attribute()
def __init__(self, d):
super(AnalyticFlow, self).__init__(d=d)
self.flow_ = abstract_attribute()
def __init__(self, *args, verbosity=None, **kwargs):
self.model_trainer = abstract_attribute()
self.logger = logging.getLogger(__name__).getChild(
self.__class__.__name__ + ":" + hex(id(self)))
self.set_verbosity(verbosity)
def __init__(self, *, d):
super(GeneralBackpropJacobianFlow, self).__init__(d=d)
# A backprop Jacobian flow must implement an attribute flow_
# which computes the transformation in a differentiable way
self.flow_ = abstract_attribute()