ray.rllib.utils.exploration.epsilon_greedy.EpsilonGreedy.get_exploration_optimizer

EpsilonGreedy.get_exploration_optimizer(optimizers: List[torch.optim.Optimizer | tf.keras.optimizers.Optimizer]) → List[torch.optim.Optimizer | tf.keras.optimizers.Optimizer]

May add optimizer(s) to the Policy’s own optimizers.

The number of optimizers (Policy’s plus Exploration’s optimizers) must match the number of loss terms produced by the Policy’s loss function and the Exploration component’s loss terms.

Parameters:

optimizers – The list of the Policy’s local optimizers.

Returns:

The updated list of local optimizers to use on the different loss terms.