Ray AIR Session

See also

See this Ray Train user guide and this Ray Tune user guide for usage examples of ray.air.session in the respective libraries.

Report Metrics and Save Checkpoints

session.report(metrics, *[, checkpoint])

Report metrics and optionally save a checkpoint.

Retrieve Checkpoints and Datasets

session.get_checkpoint()	Access the session's last checkpoint to resume from if applicable.
session.get_dataset_shard([dataset_name])	Returns the ray.data.DatasetIterator shard for this worker.

AIR Session Metadata

session.get_experiment_name()	Experiment name for the corresponding trial.
session.get_trial_name()	Trial name for the corresponding trial.
session.get_trial_id()	Trial id for the corresponding trial.
session.get_trial_resources()	Trial resources for the corresponding trial.
session.get_trial_dir()	Log directory corresponding to the trial directory for a Tune session.
session.get_world_size()	Get the current world size (i.e.
session.get_world_rank()	Get the world rank of this worker.
session.get_local_world_size()	Get the local rank of this worker (rank of the worker on its node).
session.get_local_rank()	Get the local rank of this worker (rank of the worker on its node).
session.get_node_rank()	Get the local rank of this worker (rank of the worker on its node).