Tianshou/examples/box2d/acrobot_dualdqn.py

import argparse
import os
import pprint

import gymnasium as gym
import numpy as np
import torch
from torch.utils.tensorboard import SummaryWriter

from tianshou.data import Collector, VectorReplayBuffer
from tianshou.env import DummyVectorEnv
from tianshou.policy import DQNPolicy
from tianshou.trainer import OffpolicyTrainer
from tianshou.utils import TensorboardLogger
from tianshou.utils.net.common import Net


def get_args():
    parser = argparse.ArgumentParser()
    parser.add_argument("--task", type=str, default="Acrobot-v1")
    parser.add_argument("--seed", type=int, default=0)
    parser.add_argument("--eps-test", type=float, default=0.05)
    parser.add_argument("--eps-train", type=float, default=0.5)
    parser.add_argument("--buffer-size", type=int, default=20000)
    parser.add_argument("--lr", type=float, default=1e-3)
    parser.add_argument("--gamma", type=float, default=0.95)
    parser.add_argument("--n-step", type=int, default=3)
    parser.add_argument("--target-update-freq", type=int, default=320)
    parser.add_argument("--epoch", type=int, default=10)
    parser.add_argument("--step-per-epoch", type=int, default=100000)
    parser.add_argument("--step-per-collect", type=int, default=100)
    parser.add_argument("--update-per-step", type=float, default=0.01)
    parser.add_argument("--batch-size", type=int, default=64)
    parser.add_argument("--hidden-sizes", type=int, nargs="*", default=[128])
    parser.add_argument("--dueling-q-hidden-sizes", type=int, nargs="*", default=[128, 128])
    parser.add_argument("--dueling-v-hidden-sizes", type=int, nargs="*", default=[128, 128])
    parser.add_argument("--training-num", type=int, default=10)
    parser.add_argument("--test-num", type=int, default=100)
    parser.add_argument("--logdir", type=str, default="log")
    parser.add_argument("--render", type=float, default=0.0)
    parser.add_argument(
        "--device",
        type=str,
        default="cuda" if torch.cuda.is_available() else "cpu",
    )
    return parser.parse_args()


def test_dqn(args=get_args()):
    env = gym.make(args.task)
    args.state_shape = env.observation_space.shape or env.observation_space.n
    args.action_shape = env.action_space.shape or env.action_space.n
    # train_envs = gym.make(args.task)
    # you can also use tianshou.env.SubprocVectorEnv
    train_envs = DummyVectorEnv([lambda: gym.make(args.task) for _ in range(args.training_num)])
    # test_envs = gym.make(args.task)
    test_envs = DummyVectorEnv([lambda: gym.make(args.task) for _ in range(args.test_num)])
    # seed
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    train_envs.seed(args.seed)
    test_envs.seed(args.seed)
    # model
    Q_param = {"hidden_sizes": args.dueling_q_hidden_sizes}
    V_param = {"hidden_sizes": args.dueling_v_hidden_sizes}
    net = Net(
        args.state_shape,
        args.action_shape,
        hidden_sizes=args.hidden_sizes,
        device=args.device,
        dueling_param=(Q_param, V_param),
    ).to(args.device)
    optim = torch.optim.Adam(net.parameters(), lr=args.lr)
    policy = DQNPolicy(
        model=net,
        optim=optim,
        action_space=env.action_space,
        discount_factor=args.gamma,
        estimation_step=args.n_step,
        target_update_freq=args.target_update_freq,
    )
    # collector
    train_collector = Collector(
        policy,
        train_envs,
        VectorReplayBuffer(args.buffer_size, len(train_envs)),
        exploration_noise=True,
    )
    test_collector = Collector(policy, test_envs, exploration_noise=True)
    # policy.set_eps(1)
    train_collector.collect(n_step=args.batch_size * args.training_num)
    # log
    log_path = os.path.join(args.logdir, args.task, "dqn")
    writer = SummaryWriter(log_path)
    logger = TensorboardLogger(writer)

    def save_best_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, "policy.pth"))

    def stop_fn(mean_rewards):
        return mean_rewards >= env.spec.reward_threshold

    def train_fn(epoch, env_step):
        if env_step <= 100000:
            policy.set_eps(args.eps_train)
        elif env_step <= 500000:
            eps = args.eps_train - (env_step - 100000) / 400000 * (0.5 * args.eps_train)
            policy.set_eps(eps)
        else:
            policy.set_eps(0.5 * args.eps_train)

    def test_fn(epoch, env_step):
        policy.set_eps(args.eps_test)

    # trainer
    result = OffpolicyTrainer(
        policy=policy,
        train_collector=train_collector,
        test_collector=test_collector,
        max_epoch=args.epoch,
        step_per_epoch=args.step_per_epoch,
        step_per_collect=args.step_per_collect,
        episode_per_test=args.test_num,
        batch_size=args.batch_size,
        update_per_step=args.update_per_step,
        train_fn=train_fn,
        test_fn=test_fn,
        stop_fn=stop_fn,
        save_best_fn=save_best_fn,
        logger=logger,
    ).run()

    assert stop_fn(result["best_reward"])
    if __name__ == "__main__":
        pprint.pprint(result)
        # Let's watch its performance!
        policy.eval()
        policy.set_eps(args.eps_test)
        test_envs.seed(args.seed)
        test_collector.reset()
        result = test_collector.collect(n_episode=args.test_num, render=args.render)
        rews, lens = result["rews"], result["lens"]
        print(f"Final reward: {rews.mean()}, length: {lens.mean()}")


if __name__ == "__main__":
    test_dqn(get_args())