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