:py:mod:`n2j.models.flow`
=========================

.. py:module:: n2j.models.flow

.. autoapi-nested-parse::

   Credit to Miles Cranmer:
   https://github.com/MilesCranmer/easy_normalizing_flow/blob/master/flow.py



Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   n2j.models.flow.MaskedLinear
   n2j.models.flow.MADE
   n2j.models.flow.MAF
   n2j.models.flow.Perm
   n2j.models.flow.Flow




Attributes
~~~~~~~~~~

.. autoapisummary::

   n2j.models.flow.DEBUG


.. py:data:: DEBUG
   :value: False

   

.. py:class:: MaskedLinear(in_features, out_features, bias=True)

   Bases: :py:obj:`torch.nn.Linear`

   same as Linear except has a configurable mask on the weights 

   .. py:method:: set_mask(mask)


   .. py:method:: forward(input)



.. py:class:: MADE(nin, hidden_sizes, nout, num_masks=1, natural_ordering=False)

   Bases: :py:obj:`torch.nn.Module`

   .. py:method:: update_masks()


   .. py:method:: forward(x)



.. py:class:: MAF(features, context, hidden=100, nlayers=1)

   Bases: :py:obj:`torch.nn.Module`

   x0 only depends on x0, etc

   .. py:method:: fmualpha(x)


   .. py:method:: load_context(x, context)


   .. py:method:: invert(u, context)


   .. py:method:: forward(x, context)



.. py:class:: Perm(nvars, perm=None)

   Bases: :py:obj:`torch.nn.Module`

   .. py:method:: forward(x, context)


   .. py:method:: invert(x, context)



.. py:class:: Flow(*layers)

   Bases: :py:obj:`torch.nn.Module`

   .. py:method:: forward(x, context)


   .. py:method:: invert(u, context)