BernoulliProbabilityDistribution

class maze.distributions.bernoulli.BernoulliProbabilityDistribution(*args, **kwds)

Bernoulli Torch probability distribution for multi-binary action spaces.

Parameters

• logits – the action selection logits.

• action_space – The gym action space.

• temperature – The distribution temperature parameter.

deterministic_sample() → torch.Tensor

(overrides TorchProbabilityDistribution)

implementation of ProbabilityDistribution interface

entropy() → torch.Tensor

(overrides TorchProbabilityDistribution)

implementation of ProbabilityDistribution interface

kl(other: maze.distributions.torch_dist.TorchProbabilityDistribution) → torch.Tensor

(overrides TorchProbabilityDistribution)

implementation of ProbabilityDistribution interface

log_prob(actions: torch.Tensor) → torch.Tensor

(overrides TorchProbabilityDistribution)

implementation of ProbabilityDistribution interface

classmethod required_logits_shape(action_space: gym.spaces.MultiBinary) → Sequence[int]

(overrides TorchProbabilityDistribution)

implementation of TorchProbabilityDistribution interface