def dimwise(transforms):
net = nn.Sequential(DenseTransformer(d_input=D//2, d_output=P*D//2, d_model=args.d_model, nhead=args.nhead,
num_layers=args.num_layers, dim_feedforward=4*args.d_model,
dropout=args.dropout, activation=args.activation,
checkpoint_blocks=args.checkpoint_blocks),
ElementwiseParams1d(P))
if args.affine: transforms.append(AffineCouplingBijection(net, split_dim=1, scale_fn=scale_fn(args.scale_fn)))
else: transforms.append(AdditiveCouplingBijection(net, split_dim=1))
transforms.append(Reverse(D, dim=1))
return transforms
def lenwise(transforms):
net = nn.Sequential(
DenseTransformer(d_input=D,
d_output=P * D,
d_model=args.d_model,
nhead=args.nhead,
num_layers=args.num_layers,
dim_feedforward=4 * args.d_model,
dropout=args.dropout,
activation=args.activation,
checkpoint_blocks=args.checkpoint_blocks),
ElementwiseParams1d(P))
if args.affine:
transforms.append(
AffineCouplingBijection(net,
split_dim=2,
scale_fn=scale_fn(args.scale_fn)))
else:
transforms.append(AdditiveCouplingBijection(net, split_dim=2))
if args.stochperm: transforms.append(StochasticPermutation(dim=2))
else: transforms.append(Shuffle(L, dim=2))
return transforms
D = 2 # Number of data dimensions
A = 2 + args.augdim # Number of augmented data dimensions
P = 2 if args.affine else 1 # Number of elementwise parameters
transforms = [Augment(StandardNormal((args.augdim, )), x_size=D)]
for _ in range(args.num_flows):
net = nn.Sequential(
MLP(A // 2,
P * A // 2,
hidden_units=args.hidden_units,
activation=args.activation), ElementwiseParams(P))
if args.affine:
transforms.append(
AffineCouplingBijection(net, scale_fn=scale_fn(args.scale_fn)))
else:
transforms.append(AdditiveCouplingBijection(net))
if args.actnorm: transforms.append(ActNormBijection(D))
transforms.append(Reverse(A))
transforms.pop()
model = Flow(base_dist=StandardNormal((A, )),
transforms=transforms).to(args.device)
#######################
## Specify optimizer ##
#######################
if args.optimizer == 'adam':
optimizer = Adam(model.parameters(), lr=args.lr)
elif args.optimizer == 'adamax':
optimizer = Adamax(model.parameters(), lr=args.lr)
else: transforms.append(ConditionalAdditiveCouplingBijection(coupling_net=net, context_net=context_net, num_condition=I))
if args.actnorm: transforms.append(ActNormBijection(D))
if args.permutation == 'reverse': transforms.append(Reverse(D))
elif args.permutation == 'shuffle': transforms.append(Shuffle(D))
transforms.pop()
decoder = ConditionalFlow(base_dist=StandardNormal((D,)), transforms=transforms).to(args.device)
# Flow
transforms = []
for _ in range(args.num_flows):
net = nn.Sequential(MLP(I, P*O,
hidden_units=args.hidden_units,
activation=args.activation),
ElementwiseParams(P))
if args.affine: transforms.append(AffineCouplingBijection(net, scale_fn=scale_fn(args.scale_fn), num_condition=I))
else: transforms.append(AdditiveCouplingBijection(net, num_condition=I))
if args.actnorm: transforms.append(ActNormBijection(D))
if args.permutation == 'reverse': transforms.append(Reverse(D))
elif args.permutation == 'shuffle': transforms.append(Shuffle(D))
transforms.pop()
if args.num_bits is not None:
transforms.append(Sigmoid())
transforms.append(VariationalQuantization(decoder, num_bits=args.num_bits))
pi = Flow(base_dist=target,
transforms=transforms).to(args.device)
p = StandardNormal(shape).to(args.device)
#######################
def __init__(self, num_flows, actnorm, affine, scale_fn_str, hidden_units,
activation, range_flow, augment_size, base_dist):
D = 2 # Number of data dimensions
if base_dist == "uniform":
classifier = MLP(D,
D // 2,
hidden_units=hidden_units,
activation=activation,
out_lambda=lambda x: x.view(-1))
transforms = [
ElementAbsSurjection(classifier=classifier),
ShiftBijection(shift=torch.tensor([[0.0, 4.0]])),
ScaleBijection(scale=torch.tensor([[1 / 4, 1 / 8]]))
]
base = StandardUniform((D, ))
else:
A = D + augment_size # Number of augmented data dimensions
P = 2 if affine else 1 # Number of elementwise parameters
# initialize flow with either augmentation or Abs surjection
if augment_size > 0:
assert augment_size % 2 == 0
transforms = [
Augment(StandardNormal((augment_size, )), x_size=D)
]
else:
transforms = [SimpleAbsSurjection()]
if range_flow == 'logit':
transforms += [
ScaleBijection(scale=torch.tensor([[1 / 4, 1 / 4]])),
Logit()
]
elif range_flow == 'softplus':
transforms += [SoftplusInverse()]
# apply coupling layer flows
for _ in range(num_flows):
net = nn.Sequential(
MLP(A // 2,
P * A // 2,
hidden_units=hidden_units,
activation=activation), ElementwiseParams(P))
if affine:
transforms.append(
AffineCouplingBijection(
net, scale_fn=scale_fn(scale_fn_str)))
else:
transforms.append(AdditiveCouplingBijection(net))
if actnorm: transforms.append(ActNormBijection(D))
transforms.append(Reverse(A))
transforms.pop()
base = StandardNormal((A, ))
super(UnconditionalFlow, self).__init__(base_dist=base,
transforms=transforms)