two-step-game-qmix-without-mixer:
    env: ray.rllib.examples.env.two_step_game.TwoStepGameWithGroupedAgents
    run: QMIX
    stop:
        episode_reward_mean: 7.0
        timesteps_total: 70000
    config:
        # QMIX only supports torch for now.
        framework: torch

        env_config:
            env_config:
              separate_state_space: true
              one_hot_state_encoding: true

        exploration_config:
            final_epsilon: 0.0

        rollout_fragment_length: 4
        train_batch_size: 32
        num_workers: 0
        mixer: null