# RL-Games PPO Configuration for Long Trajectory Assembly
# Copyright (c) 2022-2025, The Isaac Lab Project Developers.

params:
  seed: 42

  # Environment wrapper clipping
  env:
    clip_observations: 5.0
    clip_actions: 1.0

  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
        fixed_sigma: True

    mlp:
      units: [512, 256, 128]
      activation: elu
      d2rl: False
      
      initializer:
        name: default
      regularizer:
        name: None

  load_checkpoint: False
  load_path: ''

  config:
    name: Galaxea-LongTrajectoryAssembly-Direct-v0
    full_experiment_name: LongTrajectoryAssembly
    
    env_name: rlgpu
    device: 'cuda:0'
    device_name: 'cuda:0'
    multi_gpu: False
    ppo: True
    mixed_precision: False
    normalize_input: True
    normalize_value: True
    # value_bootstrap: True  # Commented out to match Isaac Lab examples
    num_actors: -1  # Will be set by num_envs
    reward_shaper:
      scale_value: 1.0
    normalize_advantage: True
    
    gamma: 0.99
    tau: 0.95
    learning_rate: 3e-4
    lr_schedule: adaptive
    kl_threshold: 0.008
    
    score_to_win: 100000
    max_epochs: 5000
    save_best_after: 100
    save_frequency: 100
    print_stats: True
    
    grad_norm: 1.0
    entropy_coef: 0.001
    truncate_grads: True
    
    e_clip: 0.2
    clip_value: True
    
    # PPO specific
    horizon_length: 32
    minibatch_size: 16384
    mini_epochs: 8
    critic_coef: 2
    bounds_loss_coef: 0.0001  # Add bounds loss coefficient to prevent b_loss error
    
    # Training
    games_to_track: 100
    player:
      deterministic: True
      games_num: 1000000
      print_stats: True