def create_base_transform(i):
if args.base_transform_type == 'rq':
return transforms.PiecewiseRationalQuadraticCouplingTransform(
mask=utils.create_alternating_binary_mask(
features=dim,
even=(i % 2 == 0)
),
transform_net_create_fn=lambda in_features, out_features:
nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
num_blocks=args.num_transform_blocks,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
),
num_bins=args.num_bins,
apply_unconditional_transform=False,
)
elif args.base_transform_type == 'affine':
return transforms.AffineCouplingTransform(
mask=utils.create_alternating_binary_mask(
features=dim,
even=(i % 2 == 0)
),
transform_net_create_fn=lambda in_features, out_features:
nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
num_blocks=args.num_transform_blocks,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
)
)
elif args.base_transform_type == 'rl':
return transforms.PiecewiseRationalLinearCouplingTransform(
mask=utils.create_alternating_binary_mask(
features=dim,
even=(i % 2 == 0)
),
transform_net_create_fn=lambda in_features, out_features:
nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
num_blocks=args.num_transform_blocks,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
),
num_bins=args.num_bins,
apply_unconditional_transform=False,
)
else:
raise ValueError
def create_resnet(in_features, out_features):
return nn_.ResidualNet(
in_features,
out_features,
hidden_features=hidden_features,
num_blocks=num_blocks_per_layer,
activation=activation,
dropout_probability=dropout_probability,
use_batch_norm=batch_norm_within_layers
)
# num_blocks=2,
# use_batch_norm=True
# ),
# num_bins=args.num_bins,
# apply_unconditional_transform=False
# )
# ])
transform = transforms.CompositeTransform([
# transforms.Sigmoid(),
transforms.PiecewiseRationalLinearCouplingTransform(
mask=utils.create_alternating_binary_mask(features=dim, even=True),
transform_net_create_fn=lambda in_features, out_features: nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=32,
num_blocks=2,
use_batch_norm=True
),
tails='linear',
tail_bound=5,
num_bins=args.num_bins,
apply_unconditional_transform=False
),
transforms.PiecewiseRationalLinearCouplingTransform(
mask=utils.create_alternating_binary_mask(features=dim, even=False),
transform_net_create_fn=lambda in_features, out_features: nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=32,
num_blocks=2,
def create_net(in_features, out_features):
return nn_.ResidualNet(in_features,
out_features,
hidden_features=30,
num_blocks=5)
def create_base_transform(i):
if args.base_transform_type == 'affine-coupling':
return transforms.AffineCouplingTransform(
mask=utils.create_alternating_binary_mask(features, even=(i % 2 == 0)),
transform_net_create_fn=lambda in_features, out_features: nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
context_features=None,
num_blocks=args.num_transform_blocks,
activation=F.relu,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
)
)
elif args.base_transform_type == 'quadratic-coupling':
return transforms.PiecewiseQuadraticCouplingTransform(
mask=utils.create_alternating_binary_mask(features, even=(i % 2 == 0)),
transform_net_create_fn=lambda in_features, out_features: nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
context_features=None,
num_blocks=args.num_transform_blocks,
activation=F.relu,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
),
num_bins=args.num_bins,
tails='linear',
tail_bound=args.tail_bound,
apply_unconditional_transform=args.apply_unconditional_transform
)
elif args.base_transform_type == 'rq-coupling':
return transforms.PiecewiseRationalQuadraticCouplingTransform(
mask=utils.create_alternating_binary_mask(features, even=(i % 2 == 0)),
transform_net_create_fn=lambda in_features, out_features: nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
context_features=None,
num_blocks=args.num_transform_blocks,
activation=F.relu,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
),
num_bins=args.num_bins,
tails='linear',
tail_bound=args.tail_bound,
apply_unconditional_transform=args.apply_unconditional_transform
)
elif args.base_transform_type == 'affine-autoregressive':
return transforms.MaskedAffineAutoregressiveTransform(
features=features,
hidden_features=args.hidden_features,
context_features=None,
num_blocks=args.num_transform_blocks,
use_residual_blocks=True,
random_mask=False,
activation=F.relu,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
)
elif args.base_transform_type == 'quadratic-autoregressive':
return transforms.MaskedPiecewiseQuadraticAutoregressiveTransform(
features=features,
hidden_features=args.hidden_features,
context_features=None,
num_bins=args.num_bins,
tails='linear',
tail_bound=args.tail_bound,
num_blocks=args.num_transform_blocks,
use_residual_blocks=True,
random_mask=False,
activation=F.relu,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
)
elif args.base_transform_type == 'rq-autoregressive':
return transforms.MaskedPiecewiseRationalQuadraticAutoregressiveTransform(
features=features,
hidden_features=args.hidden_features,
context_features=None,
num_bins=args.num_bins,
tails='linear',
tail_bound=args.tail_bound,
num_blocks=args.num_transform_blocks,
use_residual_blocks=True,
random_mask=False,
activation=F.relu,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
)
else:
raise ValueError
distribution = distributions.TweakedUniform(
low=torch.zeros(dim),
high=torch.ones(dim)
)
transform = transforms.CompositeTransform([
transforms.CompositeTransform([
transforms.PiecewiseRationalQuadraticCouplingTransform(
mask=utils.create_alternating_binary_mask(
features=dim,
even=(i % 2 == 0)
),
transform_net_create_fn=lambda in_features, out_features:
nn_.ResidualNet(
in_features=in_features,
out_features=out_features,
hidden_features=args.hidden_features,
num_blocks=args.num_transform_blocks,
dropout_probability=args.dropout_probability,
use_batch_norm=args.use_batch_norm
),
num_bins=args.num_bins,
tails=None,
tail_bound=1,
# apply_unconditional_transform=args.apply_unconditional_transform,
min_bin_width=args.min_bin_width
),
]) for i in range(args.num_flow_steps)
])
flow = flows.Flow(transform, distribution).to(device)
path = os.path.join(cutils.get_final_root(), '{}-final.t'.format(dataset_name))
state_dict = torch.load(path)