""" scripts/train.py — Phase 3 RL training scaffold ------------------------------------------------- Trains PPO on PrecisionIrrigation-v0 using Stable-Baselines3. Usage: python scripts/train.py python scripts/train.py --timesteps 200000 --algo dqn Requires: pip install stable-baselines3[extra] tensorboard """ import sys import os import argparse sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) try: import gymnasium as gym import envs # noqa: F401 — triggers registration except ImportError: print("āŒ gymnasium not installed. Run: pip install gymnasium") sys.exit(1) try: from stable_baselines3 import PPO, DQN from stable_baselines3.common.env_util import make_vec_env from stable_baselines3.common.evaluation import evaluate_policy from stable_baselines3.common.callbacks import EvalCallback, StopTrainingOnRewardThreshold except ImportError: print("āŒ stable-baselines3 not installed. Run: pip install stable-baselines3[extra]") sys.exit(1) ALGO_MAP = {"ppo": PPO, "dqn": DQN} MODELS_DIR = os.path.join(os.path.dirname(__file__), "..", "models") LOGS_DIR = os.path.join(os.path.dirname(__file__), "..", "logs") os.makedirs(MODELS_DIR, exist_ok=True) os.makedirs(LOGS_DIR, exist_ok=True) def train(algo_name: str = "ppo", total_timesteps: int = 100_000, n_envs: int = 4): AlgoCls = ALGO_MAP[algo_name.lower()] print(f"\n🚜 Training {algo_name.upper()} on PrecisionIrrigation-v0") print(f" timesteps={total_timesteps:,} parallel_envs={n_envs}\n") # Vectorised environments for faster training vec_env = make_vec_env("PrecisionIrrigation-v0", n_envs=n_envs, seed=0) eval_env = gym.make("PrecisionIrrigation-v0") # Stop early if reward threshold reached stop_cb = StopTrainingOnRewardThreshold(reward_threshold=25.0, verbose=1) eval_cb = EvalCallback( eval_env, best_model_save_path=MODELS_DIR, log_path=LOGS_DIR, eval_freq=max(total_timesteps // 20, 1000), n_eval_episodes=10, callback_on_new_best=stop_cb, verbose=1, ) model_kwargs: dict = {"verbose": 1, "tensorboard_log": LOGS_DIR} if algo_name == "ppo": model_kwargs.update({"n_steps": 256, "batch_size": 64, "learning_rate": 3e-4}) model = AlgoCls("MultiInputPolicy", vec_env, **model_kwargs) model.learn(total_timesteps=total_timesteps, callback=eval_cb, progress_bar=True) save_path = os.path.join(MODELS_DIR, f"{algo_name}_irrigation_final") model.save(save_path) print(f"\nšŸ’¾ Model saved → {save_path}.zip") # Final evaluation mean_r, std_r = evaluate_policy(model, eval_env, n_eval_episodes=20, deterministic=True) print(f"šŸ“Š Final eval — mean reward: {mean_r:.2f} ± {std_r:.2f}") vec_env.close() eval_env.close() return model if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument("--algo", default="ppo", choices=["ppo", "dqn"]) parser.add_argument("--timesteps", default=100_000, type=int) parser.add_argument("--n-envs", default=4, type=int) args = parser.parse_args() train(args.algo, args.timesteps, args.n_envs)