get_logger

torchrl.record.loggers.get_logger(logger_type: str, logger_name: str, experiment_name: str, **kwargs) → Logger

Get a logger instance of the provided logger_type.

Parameters:

• logger_type (str) – One of tensorboard / csv / wandb / mlflow / trackio. If empty, None is returned.

• logger_name (str) – Name to be used as a log_dir

• experiment_name (str) – Name of the experiment

• kwargs (dict[str]) – might contain either wandb_kwargs, mlflow_kwargs or trackio_kwargs. Additionally supports use_ray_service (bool) and ray_actor_options (dict) to run the logger as a Ray actor.