lit_sb3.core.hyperparams_opt

Optuna hyperparameter samplers for all supported algorithms.

Adapted from https://github.com/DLR-RM/rl-baselines3-zoo/blob/master/rl_zoo3/hyperparams_opt.py

Requires optuna to be installed::

pip install optuna

Functions

convert_onpolicy_params — Convert sampled on-policy parameter dict to algorithm kwargs.
convert_offpolicy_params — Convert sampled off-policy parameter dict to algorithm kwargs.
convert_ars_params — Convert sampled ARS parameter dict to algorithm kwargs.
sample_a2c_params — Sample A2C hyperparameters.
sample_ppo_params — Sample PPO hyperparameters.
sample_ppo_lstm_params — Sample RecurrentPPO (PPO-LSTM) hyperparameters.
sample_trpo_params — Sample TRPO hyperparameters.
sample_sac_params — Sample SAC hyperparameters.
sample_tqc_params — Sample TQC hyperparameters (TQC = SAC + distributional RL).
sample_td3_params — Sample TD3 / DDPG hyperparameters.
sample_dqn_params — Sample DQN hyperparameters.
sample_qrdqn_params — Sample QR-DQN hyperparameters (QR-DQN = DQN + distributional RL).
sample_ars_params — Sample ARS hyperparameters.
sample_crossq_params — Sample CrossQ hyperparameters (CrossQ ≈ SAC variant).

source convert_onpolicy_params(sampled_params: dict[str, Any]) → dict[str, Any]

Convert sampled on-policy parameter dict to algorithm kwargs.

source convert_offpolicy_params(sampled_params: dict[str, Any]) → dict[str, Any]

Convert sampled off-policy parameter dict to algorithm kwargs.

source convert_ars_params(sampled_params: dict[str, Any]) → dict[str, Any]

Convert sampled ARS parameter dict to algorithm kwargs.

source sample_a2c_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample A2C hyperparameters.

source sample_ppo_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample PPO hyperparameters.

source sample_ppo_lstm_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample RecurrentPPO (PPO-LSTM) hyperparameters.

source sample_trpo_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample TRPO hyperparameters.

source sample_sac_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample SAC hyperparameters.

source sample_tqc_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample TQC hyperparameters (TQC = SAC + distributional RL).

source sample_td3_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample TD3 / DDPG hyperparameters.

source sample_dqn_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample DQN hyperparameters.

source sample_qrdqn_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample QR-DQN hyperparameters (QR-DQN = DQN + distributional RL).

source sample_ars_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample ARS hyperparameters.

source sample_crossq_params(trial: Any, n_actions: int, n_envs: int, additional_args: dict[str, Any]) → dict[str, Any]

Sample CrossQ hyperparameters (CrossQ ≈ SAC variant).