ParallelLoadedImitationDataset

class maze.train.trainers.imitation.parallel_loaded_im_data_set.ParallelLoadedImitationDataset(*args: Any, **kwargs: Any)

A version of the in-memory dataset that loads all data in parallel.

This significantly speeds up data loading in cases where conversion of MazeStates and MazeActions into actions and observations is demanding.

Parameters

• trajectory_data_dir – See the parent class.

• env_factory – See the parent class.

• n_workers – Number of worker processes to load data in.