d3rlpy.datasets.get_d4rl

d3rlpy.datasets.get_d4rl(env_name, transition_picker=None, trajectory_slicer=None, render_mode=None)

Returns d4rl dataset and envrironment.

The dataset is provided through d4rl.

from d3rlpy.datasets import get_d4rl

dataset, env = get_d4rl('hopper-medium-v0')

References

• Fu et al., D4RL: Datasets for Deep Data-Driven Reinforcement Learning.

• https://github.com/rail-berkeley/d4rl

Parameters

• env_name (str) – environment id of d4rl dataset.

• transition_picker (Optional[d3rlpy.dataset.transition_pickers.TransitionPickerProtocol]) – TransitionPickerProtocol object.

• trajectory_slicer (Optional[d3rlpy.dataset.trajectory_slicers.TrajectorySlicerProtocol]) – TrajectorySlicerProtocol object.

• render_mode (Optional[str]) – Mode of rendering (human, rgb_array).

Returns

tuple of d3rlpy.dataset.ReplayBuffer and gym environment.

Return type

Tuple[d3rlpy.dataset.replay_buffer.ReplayBuffer, gym.core.Env[numpy.ndarray[Any, numpy.dtype[Any]], numpy.ndarray[Any, numpy.dtype[Any]]]]