flatland.evaluators.trajectory_evaluator module#
- class flatland.evaluators.trajectory_evaluator.TrajectoryEvaluator(trajectory: Trajectory, callbacks: FlatlandCallbacks | None = None)[source]#
Bases:
object- evaluate(start_step: int | None = None, end_step: int | None = None, snapshot_interval=0, tqdm_kwargs: dict | None = None)[source]#
- The data is structured as follows:
- -30x30 map
Contains the data to replay the episodes. - <n>_trains – for n in 10,15,20,50
- event_logs
ActionEvents.discrete_action – holds set of action to be replayed for the related episodes. TrainMovementEvents.trains_arrived – holds success rate for the related episodes. TrainMovementEvents.trains_positions – holds the positions for the related episodes.
- serialised_state
<ep_id>.pkl – Holds the pickled environment version for the episode.
- All these episodes are with constant speed of 1 and malfunctions free.
Parameters
- end_stepint
stop evaluation at intermediate step excl.
- renderingbool
render while evaluating
- snapshot_intervalint
interval to write pkl snapshots to outputs/serialised_state subdirectory (not serialised_state subdirectory directly). 1 means at every step. 0 means never.
- tqdm_args: dict
additional kwargs for tqdm