metadata
license: bsd-3-clause
tags:
- autorl
- automl
- rl
pretty_name: ARLBench Performance Data
size_categories:
- 10K<n<100K
configs:
- config_name: ppo_landscape_data
data_files:
- landscapes/atari_battle_zone_ppo.csv
- landscapes/atari_double_dunk_ppo.csv
- landscapes/atari_phoenix_ppo.csv
- landscapes/atari_qbert_ppo.csv
- landscapes/atari_this_game_ppo.csv
- landscapes/box2d_continuous_lunar_lander_ppo.csv
- landscapes/box2d_lunar_lander_ppo.csv
- landscapes/brax_halfcheetah_ppo.csv
- landscapes/brax_hopper_ppo.csv
- landscapes/brax_ant_ppo.csv
- landscapes/brax_humanoid_ppo.csv
- landscapes/cc_acrobot_ppo.csv
- landscapes/cc_cartpole_ppo.csv
- landscapes/cc_continuous_mountain_car_ppo.csv
- landscapes/cc_mountain_car_ppo.csv
- landscapes/cc_pendulum_ppo.csv
- landscapes/minigrid_door_key_ppo.csv
- landscapes/minigrid_empty_random_ppo.csv
- landscapes/minigrid_four_rooms_ppo.csv
- landscapes/minigrid_unlock_ppo.csv
- config_name: dqn_landscape_data
data_files:
- landscapes/atari_battle_zone_dqn.csv
- landscapes/atari_double_dunk_dqn.csv
- landscapes/atari_phoenix_dqn.csv
- landscapes/atari_qbert_dqn.csv
- landscapes/atari_this_game_dqn.csv
- landscapes/minigrid_door_key_dqn.csv
- landscapes/minigrid_empty_random_dqn.csv
- landscapes/minigrid_four_rooms_dqn.csv
- landscapes/minigrid_unlock_dqn.csv
- landscapes/cc_acrobot_dqn.csv
- landscapes/cc_cartpole_dqn.csv
- landscapes/cc_mountain_car_dqn.csv
- landscapes/box2d_lunar_lander_dqn.csv
- config_name: sac_landscape_data
data_files:
- landscapes/box2d_bipedal_walker_sac.csv
- landscapes/box2d_continuous_lunar_lander_sac.csv
- landscapes/brax_halfcheetah_sac.csv
- landscapes/brax_hopper_sac.csv
- landscapes/brax_ant_sac.csv
- landscapes/brax_humanoid_sac.csv
- landscapes/cc_continuous_mountain_car_sac.csv
- landscapes/cc_pendulum_sac.csv
- config_name: ppo_optimisation_data
data_files:
- rs/atari_battle_zone_ppo.csv
- rs/atari_double_dunk_ppo.csv
- rs/atari_phoenix_ppo.csv
- rs/atari_qbert_ppo.csv
- rs/atari_this_game_ppo.csv
- rs/box2d_continuous_lunar_lander_ppo.csv
- rs/box2d_lunar_lander_ppo.csv
- rs/brax_halfcheetah_ppo.csv
- rs/brax_hopper_ppo.csv
- rs/brax_ant_ppo.csv
- rs/brax_humanoid_ppo.csv
- rs/cc_acrobot_ppo.csv
- rs/cc_cartpole_ppo.csv
- rs/cc_continuous_mountain_car_ppo.csv
- rs/cc_mountain_car_ppo.csv
- rs/cc_pendulum_ppo.csv
- rs/minigrid_door_key_ppo.csv
- rs/minigrid_empty_random_ppo.csv
- rs/minigrid_four_rooms_ppo.csv
- rs/minigrid_unlock_ppo.csv
- smac/atari_battle_zone_ppo.csv
- smac/atari_double_dunk_ppo.csv
- smac/atari_phoenix_ppo.csv
- smac/atari_qbert_ppo.csv
- smac/atari_this_game_ppo.csv
- smac/box2d_continuous_lunar_lander_ppo.csv
- smac/box2d_lunar_lander_ppo.csv
- smac/brax_halfcheetah_ppo.csv
- smac/brax_hopper_ppo.csv
- smac/brax_ant_ppo.csv
- smac/brax_humanoid_ppo.csv
- smac/cc_acrobot_ppo.csv
- smac/cc_cartpole_ppo.csv
- smac/cc_continuous_mountain_car_ppo.csv
- smac/cc_mountain_car_ppo.csv
- smac/cc_pendulum_ppo.csv
- smac/minigrid_door_key_ppo.csv
- smac/minigrid_empty_random_ppo.csv
- smac/minigrid_four_rooms_ppo.csv
- smac/minigrid_unlock_ppo.csv
- smac_mf/atari_battle_zone_ppo.csv
- smac_mf/atari_double_dunk_ppo.csv
- smac_mf/atari_phoenix_ppo.csv
- smac_mf/atari_qbert_ppo.csv
- smac_mf/atari_this_game_ppo.csv
- smac_mf/box2d_continuous_lunar_lander_ppo.csv
- smac_mf/box2d_lunar_lander_ppo.csv
- smac_mf/brax_halfcheetah_ppo.csv
- smac_mf/brax_hopper_ppo.csv
- smac_mf/brax_ant_ppo.csv
- smac_mf/brax_humanoid_ppo.csv
- smac_mf/cc_acrobot_ppo.csv
- smac_mf/cc_cartpole_ppo.csv
- smac_mf/cc_continuous_mountain_car_ppo.csv
- smac_mf/cc_mountain_car_ppo.csv
- smac_mf/cc_pendulum_ppo.csv
- smac_mf/minigrid_door_key_ppo.csv
- smac_mf/minigrid_empty_random_ppo.csv
- smac_mf/minigrid_four_rooms_ppo.csv
- smac_mf/minigrid_unlock_ppo.csv
- pbt/atari_battle_zone_ppo.csv
- pbt/atari_double_dunk_ppo.csv
- pbt/atari_phoenix_ppo.csv
- pbt/atari_qbert_ppo.csv
- pbt/atari_this_game_ppo.csv
- pbt/box2d_continuous_lunar_lander_ppo.csv
- pbt/box2d_lunar_lander_ppo.csv
- pbt/brax_halfcheetah_ppo.csv
- pbt/brax_hopper_ppo.csv
- pbt/brax_ant_ppo.csv
- pbt/brax_humanoid_ppo.csv
- pbt/cc_acrobot_ppo.csv
- pbt/cc_cartpole_ppo.csv
- pbt/cc_continuous_mountain_car_ppo.csv
- pbt/cc_mountain_car_ppo.csv
- pbt/cc_pendulum_ppo.csv
- pbt/minigrid_door_key_ppo.csv
- pbt/minigrid_empty_random_ppo.csv
- pbt/minigrid_four_rooms_ppo.csv
- pbt/minigrid_unlock_ppo.csv
- config_name: dqn_optimisation_data
data_files:
- rs/atari_battle_zone_dqn.csv
- rs/atari_double_dunk_dqn.csv
- rs/atari_phoenix_dqn.csv
- rs/atari_qbert_dqn.csv
- rs/atari_this_game_dqn.csv
- rs/minigrid_door_key_dqn.csv
- rs/minigrid_empty_random_dqn.csv
- rs/minigrid_four_rooms_dqn.csv
- rs/minigrid_unlock_dqn.csv
- rs/cc_acrobot_dqn.csv
- rs/cc_cartpole_dqn.csv
- rs/cc_mountain_car_dqn.csv
- rs/box2d_lunar_lander_dqn.csv
- smac/atari_battle_zone_dqn.csv
- smac/atari_double_dunk_dqn.csv
- smac/atari_phoenix_dqn.csv
- smac/atari_qbert_dqn.csv
- smac/atari_this_game_dqn.csv
- smac/minigrid_door_key_dqn.csv
- smac/minigrid_empty_random_dqn.csv
- smac/minigrid_four_rooms_dqn.csv
- smac/minigrid_unlock_dqn.csv
- smac/cc_acrobot_dqn.csv
- smac/cc_cartpole_dqn.csv
- smac/cc_mountain_car_dqn.csv
- smac/box2d_lunar_lander_dqn.csv
- smac_mf/atari_battle_zone_dqn.csv
- smac_mf/atari_double_dunk_dqn.csv
- smac_mf/atari_phoenix_dqn.csv
- smac_mf/atari_qbert_dqn.csv
- smac_mf/atari_this_game_dqn.csv
- smac_mf/minigrid_door_key_dqn.csv
- smac_mf/minigrid_empty_random_dqn.csv
- smac_mf/minigrid_four_rooms_dqn.csv
- smac_mf/minigrid_unlock_dqn.csv
- smac_mf/cc_acrobot_dqn.csv
- smac_mf/cc_cartpole_dqn.csv
- smac_mf/cc_mountain_car_dqn.csv
- smac_mf/box2d_lunar_lander_dqn.csv
- pbt/atari_battle_zone_dqn.csv
- pbt/atari_double_dunk_dqn.csv
- pbt/atari_phoenix_dqn.csv
- pbt/atari_qbert_dqn.csv
- pbt/atari_this_game_dqn.csv
- pbt/minigrid_door_key_dqn.csv
- pbt/minigrid_empty_random_dqn.csv
- pbt/minigrid_four_rooms_dqn.csv
- pbt/minigrid_unlock_dqn.csv
- pbt/cc_acrobot_dqn.csv
- pbt/cc_cartpole_dqn.csv
- pbt/cc_mountain_car_dqn.csv
- pbt/box2d_lunar_lander_dqn.csv
- config_name: sac_optimisation_data
data_files:
- rs/box2d_bipedal_walker_sac.csv
- rs/box2d_continuous_lunar_lander_sac.csv
- rs/brax_halfcheetah_sac.csv
- rs/brax_hopper_sac.csv
- rs/brax_ant_sac.csv
- rs/brax_humanoid_sac.csv
- rs/cc_continuous_mountain_car_sac.csv
- rs/cc_pendulum_sac.csv
- smac/box2d_bipedal_walker_sac.csv
- smac/box2d_continuous_lunar_lander_sac.csv
- smac/brax_halfcheetah_sac.csv
- smac/brax_hopper_sac.csv
- smac/brax_ant_sac.csv
- smac/brax_humanoid_sac.csv
- smac/cc_continuous_mountain_car_sac.csv
- smac/cc_pendulum_sac.csv
- smac_mf/box2d_bipedal_walker_sac.csv
- smac_mf/box2d_continuous_lunar_lander_sac.csv
- smac_mf/brax_halfcheetah_sac.csv
- smac_mf/brax_hopper_sac.csv
- smac_mf/brax_ant_sac.csv
- smac_mf/brax_humanoid_sac.csv
- smac_mf/cc_continuous_mountain_car_sac.csv
- smac_mf/cc_pendulum_sac.csv
- pbt/box2d_bipedal_walker_sac.csv
- pbt/box2d_continuous_lunar_lander_sac.csv
- pbt/brax_halfcheetah_sac.csv
- pbt/brax_hopper_sac.csv
- pbt/brax_ant_sac.csv
- pbt/brax_humanoid_sac.csv
- pbt/cc_continuous_mountain_car_sac.csv
- pbt/cc_pendulum_sac.csv
The ARLBench Performance Dataset
ARLBench is a benchmark designed for hyperparameter optimization (HPO) in Reinforcement Learning (RL). Given that we conducted several thousand runs to identify meaningful HPO test settings for RL, we have compiled these results into a dataset for future research and applications.
This dataset can be leveraged to:
- Meta-learn insights about the hyperparameter landscape in RL.
- Warm-start HPO tools by utilizing previously explored configurations.
Dataset Details
The dataset includes:
- Landscape data: 10 runs each for PPO, DQN, and SAC across:
- Atari-5 environments
- Four XLand gridworlds
- Four Brax walkers
- Five classic control environments
- Two Box2D environments
- Optimization data: 3 runs per optimization algorithm for each algorithm-environment combination, covering:
- Population-Based Training (PBT)
- SMAC
- SMAC with Multi-Fidelity
- Random Search
Dataset Mapping
The dataset follows this mapping:
For optimization runs, it additionally includes:
- Optimization seed: Differentiates between the five optimization runs per algorithm-environment pair.
- Optimization step: Tracks configurations evaluated at different steps.
Example Usage
You can find example notebooks demonstrating how to use:
For more details, refer to the ARLBench paper.