halfcheetah-mb-mpo:
    env: ray.rllib.examples.env.halfcheetah.HalfCheetahWrapper
    run: MBMPO
    stop:
        training_iteration: 500
    config:
         # Only supported in torch right now
        framework: torch
        horizon: 200
        num_envs_per_worker: 20
        inner_adaptation_steps: 1
        maml_optimizer_steps: 8
        gamma: 0.99
        lambda: 1.0
        lr: 0.001
        clip_param: 0.5
        kl_target: 0.01
        kl_coeff: 0.0000000001
        num_workers: 20
        num_gpus: 1
        inner_lr: 0.001
        clip_actions: False
        model:
            fcnet_hiddens: [32, 32]
            free_log_std: True