ISubGVQA.sampling.methods.target_aimle
======================================

.. py:module:: ISubGVQA.sampling.methods.target_aimle


Attributes
----------

.. autoapisummary::

   ISubGVQA.sampling.methods.target_aimle.logger


Classes
-------

.. autoapisummary::

   ISubGVQA.sampling.methods.target_aimle.BaseTargetDistribution
   ISubGVQA.sampling.methods.target_aimle.TargetDistribution
   ISubGVQA.sampling.methods.target_aimle.AdaptiveTargetDistribution


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

.. py:data:: logger

.. py:class:: BaseTargetDistribution

   Bases: :py:obj:`abc.ABC`


   Helper class that provides a standard way to create an ABC using
   inheritance.


   .. py:method:: params(theta: torch.Tensor, dy: Optional[torch.Tensor], _is_minimization: bool = False) -> torch.Tensor
      :abstractmethod:



   .. py:method:: process(theta: torch.Tensor, dy: torch.Tensor, gradient: torch.Tensor) -> torch.Tensor
      :abstractmethod:



.. py:class:: TargetDistribution(alpha: float = 1.0, beta: float = 1.0, do_gradient_scaling: bool = False, eps: float = 1e-07)

   Bases: :py:obj:`BaseTargetDistribution`


   Creates a generator of target distributions parameterized by :attr:`alpha` and :attr:`beta`.

   Example::

       >>> import torch
       >>> target_distribution = TargetDistribution(alpha=1.0, beta=1.0)
       >>> target_distribution.params(theta=torch.tensor([1.0]), dy=torch.tensor([1.0]))
       tensor([2.])

   Args:
       alpha (float): weight of the initial distribution parameters theta
       beta (float): weight of the downstream gradient dy
       do_gradient_scaling (bool): whether to scale the gradient by 1/λ or not


   .. py:attribute:: alpha
      :value: 1.0



   .. py:attribute:: beta
      :value: 1.0



   .. py:attribute:: do_gradient_scaling
      :value: False



   .. py:attribute:: eps
      :value: 1e-07



   .. py:method:: params(theta: torch.Tensor, dy: Optional[torch.Tensor], alpha: Optional[float] = None, beta: Optional[float] = None, _is_minimization: bool = False) -> torch.Tensor


   .. py:method:: process(theta: torch.Tensor, dy: torch.Tensor, gradient_3d: torch.Tensor) -> torch.Tensor


.. py:class:: AdaptiveTargetDistribution(initial_alpha: float = 1.0, initial_beta: float = 1.0, initial_grad_norm: float = 1.0, beta_update_step: float = 0.0001, beta_update_momentum: float = 0.0, grad_norm_decay_rate: float = 0.9, target_norm: float = 1.0)

   Bases: :py:obj:`BaseTargetDistribution`


   Helper class that provides a standard way to create an ABC using
   inheritance.


   .. py:attribute:: alpha
      :value: 1.0



   .. py:attribute:: beta
      :value: 1.0



   .. py:attribute:: grad_norm
      :value: 1.0



   .. py:attribute:: beta_update_step
      :value: 0.0001



   .. py:attribute:: beta_update_momentum
      :value: 0.0



   .. py:attribute:: previous_beta_update
      :value: 0.0



   .. py:attribute:: grad_norm_decay_rate
      :value: 0.9



   .. py:attribute:: target_norm
      :value: 1.0



   .. py:method:: _perturbation_magnitude(theta: torch.Tensor, dy: Optional[torch.Tensor])


   .. py:method:: params(theta: torch.Tensor, dy: Optional[torch.Tensor], _is_minimization: bool = False) -> torch.Tensor


   .. py:method:: process(theta: torch.Tensor, dy: torch.Tensor, gradient_3d: torch.Tensor) -> torch.Tensor