params/agent.yaml

params:
  seed: 0
  algo:
    name: a2c_continuous
  env:
    clip_actions: 1.0
  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
        fixed_sigma: false
    mlp:
      units:
      - 512
      - 128
      - 64
      activation: elu
      d2rl: false
      initializer:
        name: default
      regularizer:
        name: None
    rnn:
      name: lstm
      units: 1024
      layers: 2
      before_mlp: true
      concat_input: true
      layer_norm: true
  load_checkpoint: false
  load_path: ''
  config:
    name: Forge
    device: cuda:0
    full_experiment_name: gr00t/nut_thread
    env_name: rlgpu
    multi_gpu: false
    ppo: true
    mixed_precision: true
    normalize_input: true
    normalize_value: true
    value_bootstrap: true
    num_actors: 64
    reward_shaper:
      scale_value: 1.0
    normalize_advantage: true
    gamma: 0.995
    tau: 0.95
    learning_rate: 0.0001
    lr_schedule: adaptive
    schedule_type: standard
    kl_threshold: 0.008
    score_to_win: 20000
    max_epochs: 200
    save_best_after: 10
    save_frequency: 100
    print_stats: true
    grad_norm: 1.0
    entropy_coef: 0.0
    truncate_grads: true
    e_clip: 0.2
    horizon_length: 256
    minibatch_size: 512
    mini_epochs: 4
    critic_coef: 2
    clip_value: true
    seq_length: 128
    bounds_loss_coef: 0.0001
    central_value_config:
      minibatch_size: 512
      mini_epochs: 4
      learning_rate: 1e-4
      lr_schedule: adaptive
      kl_threshold: 0.008
      clip_value: true
      normalize_input: true
      truncate_grads: true
      network:
        name: actor_critic
        central_value: true
        mlp:
          units:
          - 512
          - 128
          - 64
          activation: elu
          d2rl: false
          initializer:
            name: default
          regularizer:
            name: None
        rnn:
          name: lstm
          units: 1024
          layers: 2
          before_mlp: true
          concat_input: true
          layer_norm: true
    player:
      deterministic: false
    device_name: cuda:0
    train_dir: logs/rl_games/Forge