folders:
  parent_dir: "./results/"
  model_name: "sr6_128x4_das_nc"

settings:
  game_id: "umk3"
  step_ratio: 6
  frame_shape: !!python/tuple [128, 128, 1]
  continue_game: 0.0
  action_space: "discrete"
  characters: "Skorpion"
  difficulty: 5

wrappers_settings:
  normalize_reward: true
  no_attack_buttons_combinations: true
  stack_frames: 4
  dilation: 1
  add_last_action: true
  stack_actions: 12
  scale: true
  exclude_image_scaling: true
  role_relative: true
  flatten: true
  filter_keys: ["action", "own_health", "opp_health", "own_side", "opp_side", "opp_character", "stage", "timer"]

# optuna results
# Best hyperparameters: {'gamma': 0.05944028113410932, 'max_grad_norm': 3.5407661656818026, 
# 'exponent_n_steps': 5, 'n_epochs': 14, 'batch_size': 512, 'lr': 0.014638860976621421, 
# 'ent_coef': 2.361611947920214e-06, 'clip_range': 0.3, 'gae_lambda': 0.9520674913500098, 
# 'vf_coef': 0.6420316461542878, 'net_arch': 'medium', 'activation_fn': 'leaky_relu'}

policy_kwargs:
  #net_arch: [{ pi: [64, 64], vf: [32, 32] }]
  net_arch: [256, 256]
  activation_fn: "leaky_relu"

ppo_settings:
  gamma: 0.94
  model_checkpoint: "660000"     # 0: No checkpoint, else: Load checkpoint (if previously trained)
  learning_rate: [1.0e-3, 2.5e-6] # To start
  clip_range: [0.3, 0.015] # To start
  batch_size: 512 #8 #nminibatches gave different batch size depending on the number of environments: batch_size = (n_steps * n_envs) // nminibatches
  n_epochs: 14
  n_steps: 512
  gae_lambda: 0.9520674913500098
  ent_coef: 2.361611947920214e-06
  vf_coef: 0.6420316461542878
  autosave_freq: 50000
  time_steps: 1000000