rl/configs/ppo_easy.yaml

# Phase 1 -- Masked PPO on district_backlog_easy
hyperparameters:
  learning_rate: 0.0003
  n_steps: 512
  batch_size: 64
  n_epochs: 10
  gamma: 0.99
  gae_lambda: 0.95
  clip_range: 0.2
  ent_coef: 0.01
  vf_coef: 0.5
  max_grad_norm: 0.5
  net_arch: [256, 256]

training:
  total_timesteps: 200000
  n_envs: 4
  seed: 42
  eval_freq: 16384
  n_eval_episodes: 2
  grader_eval_freq_multiplier: 4
  enable_eval_callback: true
  progress_bar: false
  model_verbose: 0
  callback_verbose: 0

target_scores:
  district_backlog_easy: 0.80