q_learning

Classes

QLearning

Helper class that provides a standard way to create an ABC using

Module Contents

class q_learning.QLearning(env: gymnasium.Env, policy: gridmind.policies.soft.q_derived.base_q_derived_soft_policy.BaseQDerivedSoftPolicy | None = None, step_size: float = 0.1, discount_factor: float = 0.9, q_initializer: str = 'zero', epsilon_decay: bool = False, epsilon: float = 0.1, summary_dir: str | None = None, write_summary: bool = True)

Bases: gridmind.algorithms.base_learning_algorithm.BaseLearningAlgorithm

Helper class that provides a standard way to create an ABC using inheritance.

num_actions

epsilon_decay = False

epsilon = 0.1

policy = None

step_size = 0.1

discount_factor = 0.9

_get_state_value_fn(force_functional_interface: bool = True)

_get_state_action_value_fn(force_functional_interface: bool = True)

_get_policy()

abstract _train_steps(num_steps: int, prediction_only: bool, *args, **kwargs)

_train_episodes(num_episodes: int, prediction_only: bool = False)

set_policy(policy: gridmind.policies.soft.q_derived.base_q_derived_soft_policy.BaseQDerivedSoftPolicy)