config.json

{
  "hidden_sizes": [
    64,
    64
  ],
  "activation": "relu",
  "actor_lr": 0.0003,
  "critic_lr": 0.0003,
  "alpha_lr": 0.0003,
  "tau": 0.005,
  "gamma": 0.97,
  "auto_entropy_tuning": true,
  "target_entropy": -16.0,
  "init_alpha": 0.2,
  "buffer_size": 10000,
  "batch_size": 64,
  "gradient_steps_per_env_step": 1,
  "warmup_steps": 100,
  "total_timesteps": 10000,
  "weight_decay": 0.0001,
  "max_grad_norm": 1.0,
  "q_value_clip": 100.0,
  "normalize_rewards": true,
  "cost_bps": 10,
  "cash_buffer": 0.02,
  "max_position_weight": 0.2,
  "reward_scale": 1.0,
  "n_stocks": 15,
  "seed": 42,
  "validation_years": 2,
  "min_cagr_improvement": 0.0,
  "sharpe_weight": 0.5,
  "sharpe_eta": 0.01,
  "training_years": 10,
  "n_eval_folds": 3
}