hopper_cql:
    env: 
        grid_search:
            #- ray.rllib.examples.env.d4rl_env.hopper_random
            - ray.rllib.examples.env.d4rl_env.hopper_medium
            #- ray.rllib.examples.env.d4rl_env.hopper_expert
            #- ray.rllib.examples.env.d4rl_env.hopper_medium_replay
    run: CQL
    config:
        # SAC Configs
        #input: d4rl.hopper-random-v0
        input: d4rl.hopper-medium-v0
        #input: d4rl.hopper-expert-v0
        #input: d4rl.hopper-medium-replay-v0
        framework: torch
        soft_horizon: False
        horizon: 1000
        Q_model:
          fcnet_activation: relu
          fcnet_hiddens: [256, 256, 256]
        policy_model:
          fcnet_activation: relu
          fcnet_hiddens: [256, 256, 256]
        tau: 0.005
        target_entropy: auto
        no_done_at_end: false
        n_step: 1
        rollout_fragment_length: 1
        prioritized_replay: false
        train_batch_size: 256
        target_network_update_freq: 0
        timesteps_per_iteration: 1000
        learning_starts: 10
        optimization:
          actor_learning_rate: 0.0001
          critic_learning_rate: 0.0003
          entropy_learning_rate: 0.0001
        num_workers: 0
        num_gpus: 1
        clip_actions: false
        normalize_actions: true
        evaluation_interval: 1
        metrics_smoothing_episodes: 5
        # CQL Configs
        min_q_weight: 5.0
        bc_iters: 20000
        temperature: 1.0
        num_actions: 10
        lagrangian: False
        evaluation_config:
            input: sampler