paint_dqn.py

from paint_ppo import main


if __name__ == '__main__':
    configuration = {
        # 'num_workers': 6,
        'num_gpus': 1,

        'double_q': True,
        'dueling': True,
        'num_atoms': 1,
        'noisy': False,
        'prioritized_replay': False,
        'n_step': 1,
        'target_network_update_freq': 1000,
        # 'lr': .0000625,
        'adam_epsilon': .00015,
        'hiddens': [512],
        'learning_starts': 200,
        'buffer_size': 5000,
        'sample_batch_size': 20,
        'train_batch_size': 32,
        'schedule_max_timesteps': 4000,
        'exploration_final_eps': 0.01,
        'exploration_fraction': 0.2,
        'prioritized_replay_alpha': 0.5,
        'beta_annealing_fraction': 1.0,
        'final_prioritized_replay_beta': 1.0,
        'timesteps_per_iteration': 200,
    }
    main('DQN', configuration)