# Runs on a single g3.4xl node
# See https://github.com/ray-project/rl-experiments for results
dueling-ddqn:
    env:
        grid_search:
            - BreakoutNoFrameskip-v4
            - BeamRiderNoFrameskip-v4
            - QbertNoFrameskip-v4
            - SpaceInvadersNoFrameskip-v4
    run: DQN
    config:
        # Works for both torch and tf.
        framework: tf
        double_q: true
        dueling: true
        num_atoms: 1
        noisy: false
        prioritized_replay: false
        n_step: 1
        target_network_update_freq: 8000
        lr: .0000625
        adam_epsilon: .00015
        hiddens: [512]
        learning_starts: 20000
        buffer_size: 1000000
        rollout_fragment_length: 4
        train_batch_size: 32
        exploration_config:
          epsilon_timesteps: 200000
          final_epsilon: 0.01
        prioritized_replay_alpha: 0.5
        final_prioritized_replay_beta: 1.0
        prioritized_replay_beta_annealing_timesteps: 2000000
        num_gpus: 0.2
        timesteps_per_iteration: 10000