List of conditioners

class torchflows.bijections.finite.autoregressive.conditioning.coupling_masks.PartialCoupling(event_shape, source_mask: ~torch.Tensor, target_mask: <module 'torch' from '/home/docs/checkouts/readthedocs.org/user_builds/torchflows/envs/latest/lib/python3.11/site-packages/torch/__init__.py'>): Coupling mask object where a part of dimensions is kept unchanged and does not affect other dimensions.

class torchflows.bijections.finite.autoregressive.conditioning.coupling_masks.Coupling(event_shape, mask: Tensor): Base object which holds coupling partition mask information.

class torchflows.bijections.finite.autoregressive.conditioning.coupling_masks.GraphicalCoupling(event_shape, edge_list: List[Tuple[int, int]])

class torchflows.bijections.finite.autoregressive.conditioning.coupling_masks.HalfSplit(event_shape)

torchflows.bijections.finite.autoregressive.conditioning.coupling_masks.make_coupling(event_shape, edge_list: List[Tuple[int, int]] = None, coupling_type: str = 'half_split', **kwargs)

Parameters:

event_shape –
coupling_type – one of [‘half_split’, ‘checkerboard’, ‘checkerboard_inverted’, ‘channel_wise’, ‘channel_wise_inverted’].
edge_list –

Returns: