rl/configs/curriculum.yaml

# Phase 2 -- Curriculum PPO across all 3 tasks
hyperparameters:
  learning_rate: 0.0002
  n_steps: 512
  batch_size: 64
  n_epochs: 10
  gamma: 0.99
  gae_lambda: 0.95
  clip_range: 0.2
  ent_coef: 0.005
  vf_coef: 0.5
  max_grad_norm: 0.5
  net_arch: [256, 256]

curriculum:
  stage1_end_frac: 0.30
  stage2_end_frac: 0.70
  stage3_weights: [0.20, 0.40, 0.40]

training:
  total_timesteps: 500000
  n_envs: 4
  seed: 42
  warm_start_from: "results/best_model/phase1_final"

target_scores:
  district_backlog_easy: 0.82
  mixed_urgency_medium: 0.72
  cross_department_hard: 0.60