CategoricalProbabilityDistribution¶
-
class
maze.distributions.categorical.CategoricalProbabilityDistribution(*args, **kwds)¶ Categorical Torch probability distribution.
- Parameters
logits – the action selection logits.
-
deterministic_sample()¶ implementation of
ProbabilityDistributioninterface
-
log_prob(actions: torch.Tensor) → torch.Tensor¶ implementation of
ProbabilityDistributioninterface
-
classmethod
required_logits_shape(action_space: gym.spaces.Discrete) → Sequence[int]¶ implementation of
TorchProbabilityDistributioninterface