runner_config.yaml

params:
  seed: 0

  algo:
    name: a2c_continuous

  model:
    name: continuous_a2c_logstd

  network:
    name: actor_critic
    separate: False

    space:
      continuous:
        mu_activation: None
        sigma_activation: None
        mu_init:
          name: default
        sigma_init:
          name: const_initializer
          val: 0. # std = 1.
        fixed_sigma: True

    mlp:
      units: [256, 128, 64]
      activation: elu
      d2rl: False

      initializer:
        name: default
      regularizer:
        name: None

    rnn:
      units: 64
      layers: 1
      name: gru
      before_mlp: True

  load_checkpoint: True # flag which sets whether to load the checkpoint
  load_path: "checkpoints" # path to the checkpoint to load

  config:
    name: default
    full_experiment_name: default
    env_name: rlgpu
    ppo: True
    mixed_precision: False
    normalize_input: True
    normalize_value: True
    value_bootstrap: True
    num_actors: 100000
    reward_shaper:
      scale_value: 1.0
    normalize_advantage: True
    gamma: 0.99
    tau: 0.95
    e_clip: 0.2
    entropy_coef: 0.0
    learning_rate: 3.e-4 # overwritten by adaptive lr_schedule
    lr_schedule: adaptive
    kl_threshold: 0.008 # target kl for adaptive lr
    truncate_grads: True
    grad_norm: 1.
    horizon_length: 200
    minibatch_size: 100000
    mini_epochs: 5
    critic_coef: 2
    clip_value: True
    seq_len: 10 # only for rnn
    bounds_loss_coef: 0.001

    max_epochs: 1000
    save_best_after: 20
    score_to_win: 20000
    save_frequency: 10
    print_stats: True

    player:
      render: False
      render_sleep: 0.
      dump_stats: True
      games_num: 1