ray.train.Result#

class ray.train.Result(metrics: Dict[str, Any] | None, checkpoint: Checkpoint | None, error: TrainingFailedError | None, path: str, metrics_dataframe: pandas.DataFrame | None = None, best_checkpoints: List[Tuple[Checkpoint, Dict[str, Any]]] | None = None, _storage_filesystem: pyarrow.fs.FileSystem | None = None, return_value: Any | None = None)#

Bases: Result

The output of trainer.fit().

metrics#: The latest set of metrics reported by the training function via ray.train.report().

checkpoint#

The latest checkpoint saved by the training function via ray.train.report().

Type:: ray.train.Checkpoint | None

return_value#

The value returned by the user-defined training function on the rank 0 worker, or None if no value was returned or if training did not complete successfully. The return value must be serializable.

Type:: Any | None

metrics_dataframe#: A DataFrame of metrics from all checkpoints saved during the run. Each row corresponds to a checkpoint.

best_checkpoints#

A list of (checkpoint, metrics) tuples for the best checkpoints saved during the run. The checkpoints retained are determined by CheckpointConfig (by default, all checkpoints are kept).

Type:: List[Tuple[ray.train.Checkpoint, Dict[str, Any]]] | None

path#: Path pointing to the run output directory on persistent storage. This can point to a remote storage location (e.g. S3) or to a local location on the head node.

error#

The execution error of the training run, if the run finished in error. This is a TrainingFailedError wrapping the original exception.

Type:: ray.train.TrainingFailedError | None

property config: Dict[str, Any] | None#: Warning

DEPRECATED: This API is deprecated and may be removed in future Ray releases.

classmethod from_path(path: str | PathLike, storage_filesystem: pyarrow.fs.FileSystem | None = None) → Result[source]#

Restore a training result from a previously saved training run path.

Parameters:

path – Path to the run output directory
storage_filesystem – Optional filesystem to use for accessing the path

Returns:

Result object with restored checkpoints and metrics

get_best_checkpoint(metric: str, mode: str) → Checkpoint | None[source]#: PublicAPI (alpha): This API is in alpha and may change before becoming stable.