"""
Launch a tuned PPO training run with safer exploration settings.
"""

import os
from datetime import datetime

import torch

from stellar.train.train_ppo import PPOTrainer


def main():
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    run_name = f"train_{timestamp}_stable_v2"
    log_dir = os.path.join("Logs", run_name)
    save_dir = os.path.join("Checkpoint", run_name)
    env_config = {
        "lambda_p": 0.05,
        "filter_cost_cap": 2500.0,
    }

    trainer = PPOTrainer(
        env_config=env_config,
        lr_actor=1e-4,
        lr_critic=3e-4,
        gamma=0.998,
        lam_gae=0.95,
        clip_eps=0.2,
        n_epochs=8,
        batch_size=256,
        n_steps=1536,
        ent_coef=0.002,
        ent_coef_final=0.0002,
        ent_anneal_episodes=200000,
        target_kl=0.01,
        device="cuda" if torch.cuda.is_available() else "cpu",
    )
    trainer.train(
        total_episodes=200000,
        log_interval=10,
        save_dir=save_dir,
        log_dir=log_dir,
        eval_interval=20,
        eval_episodes=20,
        eval_deterministic=True,
    )


if __name__ == "__main__":
    main()