n2j.models.flow#

Credit to Miles Cranmer: MilesCranmer/easy_normalizing_flow

Module Contents#

Classes#

MaskedLinear

same as Linear except has a configurable mask on the weights

MADE

MAF

x0 only depends on x0, etc

Perm

Flow

Attributes#

DEBUG

n2j.models.flow.DEBUG = False[source]#
class n2j.models.flow.MaskedLinear(in_features, out_features, bias=True)[source]#

Bases: torch.nn.Linear

same as Linear except has a configurable mask on the weights

set_mask(mask)[source]#
forward(input)[source]#
class n2j.models.flow.MADE(nin, hidden_sizes, nout, num_masks=1, natural_ordering=False)[source]#

Bases: torch.nn.Module

update_masks()[source]#
forward(x)[source]#
class n2j.models.flow.MAF(features, context, hidden=100, nlayers=1)[source]#

Bases: torch.nn.Module

x0 only depends on x0, etc

fmualpha(x)[source]#
load_context(x, context)[source]#
invert(u, context)[source]#
forward(x, context)[source]#
class n2j.models.flow.Perm(nvars, perm=None)[source]#

Bases: torch.nn.Module

forward(x, context)[source]#
invert(x, context)[source]#
class n2j.models.flow.Flow(*layers)[source]#

Bases: torch.nn.Module

forward(x, context)[source]#
invert(u, context)[source]#