flowtorch.bijectors.affineautoregressive

class

flowtorch.bijectors.AffineAutoregressive

Inherits from: flowtorch.bijectors.ops.affine.Affine, flowtorch.bijectors.autoregressive.Autoregressive

empty docstring

member

init

(self, params: Union[flowtorch.lazy.Lazy, NoneType] = None, *, shape: torch.Size, context_shape: Union[torch.Size, NoneType] = None, log_scale_min_clip: float = -5.0, log_scale_max_clip: float = 3.0, sigmoid_bias: float = 2.0) -> None

<empty docstring>

member

forward

(self, x: torch.Tensor, context: Union[torch.Tensor, NoneType] = None) -> torch.Tensor

<empty docstring>

member

forward_shape

(self, shape: torch.Size) -> torch.Size


Infers the shape of the forward computation, given the input shape.
Defaults to preserving shape.

member

inverse

(self, y: torch.Tensor, x: Union[torch.Tensor, NoneType] = None, context: Union[torch.Tensor, NoneType] = None) -> torch.Tensor

<empty docstring>

member

inverse_shape

(self, shape: torch.Size) -> torch.Size


Infers the shapes of the inverse computation, given the output shape.
Defaults to preserving shape.

member

log_abs_det_jacobian

(self, x: torch.Tensor, y: torch.Tensor, context: Union[torch.Tensor, NoneType] = None) -> torch.Tensor


Computes the log det jacobian `log |dy/dx|` given input and output.
By default, assumes a volume preserving bijection.

member

param_shapes

(self, shape: torch.Size) -> Tuple[torch.Size, torch.Size]

<empty docstring>