LogReward

class torchrl.trainers.LogReward(logname='r_training', log_pbar: bool = False, reward_key: Optional[Union[str, tuple]] = None)

Reward logger hook.

Parameters:

• logname (str, optional) – name of the rewards to be logged. Default is "r_training".

• log_pbar (bool, optional) – if True, the reward value will be logged on the progression bar. Default is False.

• reward_key (str or tuple, optional) – the key where to find the reward in the input batch. Defaults to ("next", "reward")

Examples

>>> log_reward = LogReward(("next", "reward"))
>>> trainer.register_op("pre_steps_log", log_reward)

register(trainer: Trainer, name: str = 'log_reward')

Registers the hook in the trainer at a default location.

Parameters:

• trainer (Trainer) – the trainer where the hook must be registered.

• name (str) – the name of the hook.

Note

To register the hook at another location than the default, use register_op().