ray.rllib.core.models.catalog.Catalog._get_dist_cls_from_action_space#

classmethod Catalog._get_dist_cls_from_action_space(action_space: <MagicMock name='mock.Space' id='140492860849888'>, *, framework: Optional[str] = None) → ray.rllib.models.distributions.Distribution[source]#

Returns a distribution class for the given action space.

You can get the required input dimension for the distribution by calling action_dict_cls.required_input_dim(action_space) on the retrieved class. This is useful, because the Catalog needs to find out about the required input dimension for the distribution before the model that outputs these inputs is configured.

Parameters

  • action_space – Action space of the target gym env.

  • framework – The framework to use.

Returns

The distribution class for the given action space.