ray.rllib.algorithms.algorithm_config.AlgorithmConfig.get_learner_hyperparameters

AlgorithmConfig.get_learner_hyperparameters() → ray.rllib.core.learner.learner.LearnerHyperparameters

Returns a new LearnerHyperparameters instance for the respective Learner.

The LearnerHyperparameters is a dataclass containing only those config settings from AlgorithmConfig that are used by the algorithm’s specific Learner sub-class. They allow distributing only those settings relevant for learning across a set of learner workers (instead of having to distribute the entire AlgorithmConfig object).

Note that LearnerHyperparameters should always be derived directly from a AlgorithmConfig object’s own settings and considered frozen/read-only.

Returns

A LearnerHyperparameters instance for the respective Learner.