:py:mod:`nessai.flows.utils`
============================

.. py:module:: nessai.flows.utils

.. autoapi-nested-parse::

   Various utilities for implementing normalising flows.

   ..
       !! processed by numpydoc !!


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

Classes
~~~~~~~

.. autoapisummary::

   nessai.flows.utils.MLP



Functions
~~~~~~~~~

.. autoapisummary::

   nessai.flows.utils.silu
   nessai.flows.utils.configure_model
   nessai.flows.utils.reset_weights
   nessai.flows.utils.reset_permutations
   nessai.flows.utils.create_linear_transform



Attributes
~~~~~~~~~~

.. autoapisummary::

   nessai.flows.utils.logger


.. py:data:: logger
   

   
















   ..
       !! processed by numpydoc !!

.. py:function:: silu(x)

   
   SiLU (Sigmoid-weighted Linear Unit) activation function.

   Also known as swish.

   Elfwing et al 2017: https://arxiv.org/abs/1702.03118v3















   ..
       !! processed by numpydoc !!

.. py:function:: configure_model(config)

   
   Setup the flow form a configuration dictionary.
















   ..
       !! processed by numpydoc !!

.. py:function:: reset_weights(module)

   
   Reset parameters of a given module in place.

   Uses the ``reset_parameters`` method from ``torch.nn.Module``

   Also checks the following modules from nflows

   - nflows.transforms.normalization.BatchNorm

   :Parameters:

       **module** : :obj:`torch.nn.Module`
           Module to reset














   ..
       !! processed by numpydoc !!

.. py:function:: reset_permutations(module)

   
   Resets permutations and linear transforms for a given module in place.

   Resets using the original initialisation method. This needed since they
   do not have a ``reset_parameters`` method.

   :Parameters:

       **module** : :obj:`torch.nn.Module`
           Module to reset














   ..
       !! processed by numpydoc !!

.. py:class:: MLP(in_shape, out_shape, hidden_sizes, activation=F.relu, activate_output=False)

   Bases: :py:obj:`nflows.nn.nets.MLP`

   
   MLP which can be called with context.
















   ..
       !! processed by numpydoc !!
   .. py:method:: forward(self, inputs, context=None)

      
      Forward method that allows for kwargs such as context.


      :Parameters:

          **inputs** : :obj:`torch.tensor`
              Inputs to the MLP

          **context** : None
              Conditional inputs, must be None. Only implemented to the
              function is compatible with other methods.





      :Raises:

          RuntimeError
              If the context is not None.









      ..
          !! processed by numpydoc !!


.. py:function:: create_linear_transform(linear_transform, features)

   
   Function for creating linear transforms.


   :Parameters:

       **linear_transform** : {'permutation', 'lu', 'svd'}
           Linear transform to use.

       **featres** : int
           Number of features.














   ..
       !! processed by numpydoc !!