gfn.gym.set_addition
====================

.. py:module:: gfn.gym.set_addition


Classes
-------

.. autoapisummary::

   gfn.gym.set_addition.SetAddition


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

.. py:class:: SetAddition(n_items, max_items, reward_fn, fixed_length = False, device = None, debug = False)

   Bases: :py:obj:`gfn.env.DiscreteEnv`


   Append only MDP, similarly to what is described in Remark 8 of Shen et al. 2023
   [Towards Understanding and Improving GFlowNet Training](https://proceedings.mlr.press/v202/shen23a.html)

   The state is a binary vector of length `n_items`, where 1 indicates the presence of an item.
   Actions are integers from 0 to `n_items - 1` to add the corresponding item, or `n_items` to exit.
   Adding an existing item is invalid. The trajectory must end when `max_items` are present.

   Recommended preprocessor: `IdentityPreprocessor`.

   .. attribute:: n_items

      The number of items in the set.

      :type: int

   .. attribute:: max_items

      The maximum number of items that can be added to the set.

      :type: int

   .. attribute:: reward_fn

      The reward function.

      :type: Callable

   .. attribute:: fixed_length

      Whether the trajectories have a fixed length.

      :type: bool


   .. py:attribute:: States
      :type:  type[gfn.env.DiscreteStates]


   .. py:property:: all_states
      :type: gfn.env.DiscreteStates


      Returns all the states of the environment.


   .. py:method:: backward_step(states, actions)

      Performs a backward step in the environment.

      :param states: The current states.
      :param actions: The actions to take.

      :returns: The previous states.



   .. py:attribute:: fixed_length
      :value: False



   .. py:method:: get_states_indices(states)

      Returns the indices of the states.

      :param states: The states to get the indices of.

      :returns: The indices of the states.



   .. py:method:: make_states_class()

      Returns the DiscreteStates class for the SetAddition environment.



   .. py:attribute:: max_traj_len


   .. py:attribute:: n_items


   .. py:method:: reward(final_states)

      Computes the reward for a batch of final states.

      :param final_states: The final states.

      :returns: The reward of the final states.



   .. py:attribute:: reward_fn


   .. py:method:: step(states, actions)

      Performs a step in the environment.

      :param states: The current states.
      :param actions: The actions to take.

      :returns: The next states.



   .. py:property:: terminating_states
      :type: gfn.env.DiscreteStates


      Returns the terminating states of the environment.