OakEmu-RL v0.1.0

Appearance

BehaviorCloningMinibatch ​

python
BehaviorCloningMinibatch(observations: torch.Tensor, actions: torch.Tensor, weights: torch.Tensor | None = None) -> None

A minibatch of replayed prefix steps used for behaviour cloning.

Source

Values ​

actions ​

python
actions: torch.Tensor

Recorded replay actions with shape (minibatch_size,).

Source

observations ​

python
observations: torch.Tensor

Observations with shape (minibatch_size, *obs_shape).

Source

weights ​

python
weights: torch.Tensor | None = None

Optional per-sample loss weights with shape (minibatch_size,), for example competence gating.

Source