torchrl.trainers.algorithms.configs.hooks.EarlyStoppingConfig

class torchrl.trainers.algorithms.configs.hooks.EarlyStoppingConfig(monitor: Any = 'r_evaluation', mode: str = 'max', min_delta: float = 0.0, patience: int = 100000, wait_for: int = 1000000, check_finite: bool = True, _target_: str = 'torchrl.trainers.trainers.EarlyStopping')

Configuration for the EarlyStopping hook.

Examples

>>> from torchrl.trainers.algorithms.configs.hooks import EarlyStoppingConfig
>>> from hydra.utils import instantiate
>>> hook = instantiate(
...     EarlyStoppingConfig(monitor="r_training", patience=10_000)
... )