ray.rllib.utils.exploration.epsilon_greedy.EpsilonGreedy#

class ray.rllib.utils.exploration.epsilon_greedy.EpsilonGreedy(action_space: <MagicMock name='mock.spaces.Space' id='140494124151808'>, *, framework: str, initial_epsilon: float = 1.0, final_epsilon: float = 0.05, warmup_timesteps: int = 0, epsilon_timesteps: int = 100000, epsilon_schedule: Optional[ray.rllib.utils.schedules.schedule.Schedule] = None, **kwargs)[source]#

Bases: ray.rllib.utils.exploration.exploration.Exploration

Epsilon-greedy Exploration class that produces exploration actions.

When given a Model’s output and a current epsilon value (based on some Schedule), it produces a random action (if rand(1) < eps) or uses the model-computed one (if rand(1) >= eps).

Methods

__init__(action_space, *, framework[, ...])

Create an EpsilonGreedy exploration class.

before_compute_actions(*[, timestep, ...])

Hook for preparations before policy.compute_actions() is called.

get_exploration_optimizer(optimizers)

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

on_episode_end(policy, *[, environment, ...])

Handles necessary exploration logic at the end of an episode.

on_episode_start(policy, *[, environment, ...])

Handles necessary exploration logic at the beginning of an episode.

postprocess_trajectory(policy, sample_batch)

Handles post-processing of done episode trajectories.