ray.tune.Trainable.__init__#

Trainable.__init__(config: Dict[str, Any] = None, logger_creator: Callable[[Dict[str, Any]], Logger] = None, remote_checkpoint_dir: Optional[str] = None, sync_config: Optional[ray.train.SyncConfig] = None, storage: Optional[ray.train._internal.storage.StorageContext] = None)[source]#

Initialize a Trainable.

Sets up logging and points self.logdir to a directory in which training outputs should be placed.

Subclasses should prefer defining setup() instead of overriding __init__() directly.

Parameters

  • config – Trainable-specific configuration data. By default will be saved as self.config.

  • logger_creator – (Deprecated) Function that creates a ray.tune.Logger object. If unspecified, a default logger is created.

  • remote_checkpoint_dir – Upload directory (S3 or GS path). This is per trial directory, which is different from per checkpoint directory.

  • sync_config – Configuration object for syncing. See SyncConfig.