ISubGVQA.sampling.methods.simple
================================

.. py:module:: ISubGVQA.sampling.methods.simple


Attributes
----------

.. autoapisummary::

   ISubGVQA.sampling.methods.simple.DISABLE
   ISubGVQA.sampling.methods.simple.MODE


Classes
-------

.. autoapisummary::

   ISubGVQA.sampling.methods.simple.Layer


Functions
---------

.. autoapisummary::

   ISubGVQA.sampling.methods.simple.levelwiseSL
   ISubGVQA.sampling.methods.simple.levelwiseMars
   ISubGVQA.sampling.methods.simple.log1mexp
   ISubGVQA.sampling.methods.simple.levelOrder
   ISubGVQA.sampling.methods.simple.gumbel_keys
   ISubGVQA.sampling.methods.simple.sample_subset


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

.. py:data:: DISABLE
   :value: False


.. py:data:: MODE
   :value: 'default'


.. py:function:: levelwiseSL(levels: List[torch.Tensor], idx2primesub: torch.Tensor, data: torch.Tensor, theta: torch.Tensor)

.. py:function:: levelwiseMars(levels: List[torch.Tensor], idx2primesub: torch.Tensor, data: torch.Tensor, theta: torch.Tensor, parents: torch.Tensor)

.. py:function:: log1mexp(x)

.. py:function:: levelOrder(beta)

   :type root: Node
   :rtype: List[List[int]]


.. py:function:: gumbel_keys(w, time_sampled)

.. py:function:: sample_subset(w, k, time_sampled)

   Args:
       w (Tensor): Float Tensor of weights for each element. In gumbel mode
           these are interpreted as log probabilities
       k (int): number of elements in the subset sample


.. py:class:: Layer(n, k, device, root='./simple_configs')

   .. py:attribute:: id
      :value: 0



   .. py:attribute:: parents


   .. py:attribute:: levels
      :value: []



   .. py:attribute:: true_indices


   .. py:attribute:: literal_indices


   .. py:attribute:: literal_mask


   .. py:attribute:: pos_literals


   .. py:attribute:: idx2primesub


   .. py:method:: __call__(log_probs, k)


   .. py:method:: log_pr(log_probs)


   .. py:method:: sample(lit_weights, k, time_sampled=1)