mujoco_ppo.py

#!/usr/bin/env python3

import argparse
import datetime
import os
import pprint

import numpy as np
import torch
from mujoco_env import make_mujoco_env
from torch import nn
from torch.distributions import Distribution, Independent, Normal
from torch.optim.lr_scheduler import LambdaLR

from tianshou.data import Collector, CollectStats, ReplayBuffer, VectorReplayBuffer
from tianshou.highlevel.logger import LoggerFactoryDefault
from tianshou.policy import PPOPolicy
from tianshou.policy.base import BasePolicy
from tianshou.trainer import OnpolicyTrainer
from tianshou.utils.net.common import ActorCritic, Net
from tianshou.utils.net.continuous import ActorProb, Critic


def get_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser()
    parser.add_argument("--task", type=str, default="Ant-v4")
    parser.add_argument("--seed", type=int, default=0)
    parser.add_argument("--buffer-size", type=int, default=4096)
    parser.add_argument("--hidden-sizes", type=int, nargs="*", default=[64, 64])
    parser.add_argument("--lr", type=float, default=3e-4)
    parser.add_argument("--gamma", type=float, default=0.99)
    parser.add_argument("--epoch", type=int, default=100)
    parser.add_argument("--step-per-epoch", type=int, default=30000)
    parser.add_argument("--step-per-collect", type=int, default=2048)
    parser.add_argument("--repeat-per-collect", type=int, default=10)
    parser.add_argument("--batch-size", type=int, default=64)
    parser.add_argument("--training-num", type=int, default=8)
    parser.add_argument("--test-num", type=int, default=10)
    # ppo special
    parser.add_argument("--rew-norm", type=int, default=True)
    # In theory, `vf-coef` will not make any difference if using Adam optimizer.
    parser.add_argument("--vf-coef", type=float, default=0.25)
    parser.add_argument("--ent-coef", type=float, default=0.0)
    parser.add_argument("--gae-lambda", type=float, default=0.95)
    parser.add_argument("--bound-action-method", type=str, default="clip")
    parser.add_argument("--lr-decay", type=int, default=True)
    parser.add_argument("--max-grad-norm", type=float, default=0.5)
    parser.add_argument("--eps-clip", type=float, default=0.2)
    parser.add_argument("--dual-clip", type=float, default=None)
    parser.add_argument("--value-clip", type=int, default=0)
    parser.add_argument("--norm-adv", type=int, default=0)
    parser.add_argument("--recompute-adv", type=int, default=1)
    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",
    )
    parser.add_argument("--resume-path", type=str, default=None)
    parser.add_argument("--resume-id", type=str, default=None)
    parser.add_argument(
        "--logger",
        type=str,
        default="tensorboard",
        choices=["tensorboard", "wandb"],
    )
    parser.add_argument("--wandb-project", type=str, default="mujoco.benchmark")
    parser.add_argument(
        "--watch",
        default=False,
        action="store_true",
        help="watch the play of pre-trained policy only",
    )
    return parser.parse_args()


def test_ppo(args: argparse.Namespace = get_args()) -> None:
    env, train_envs, test_envs = make_mujoco_env(
        args.task,
        args.seed,
        args.training_num,
        args.test_num,
        obs_norm=True,
    )
    args.state_shape = env.observation_space.shape or env.observation_space.n
    args.action_shape = env.action_space.shape or env.action_space.n
    args.max_action = env.action_space.high[0]
    print("Observations shape:", args.state_shape)
    print("Actions shape:", args.action_shape)
    print("Action range:", np.min(env.action_space.low), np.max(env.action_space.high))
    # seed
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    # model
    net_a = Net(
        args.state_shape,
        hidden_sizes=args.hidden_sizes,
        activation=nn.Tanh,
        device=args.device,
    )
    actor = ActorProb(
        net_a,
        args.action_shape,
        unbounded=True,
        device=args.device,
    ).to(args.device)
    net_c = Net(
        args.state_shape,
        hidden_sizes=args.hidden_sizes,
        activation=nn.Tanh,
        device=args.device,
    )
    critic = Critic(net_c, device=args.device).to(args.device)
    actor_critic = ActorCritic(actor, critic)

    torch.nn.init.constant_(actor.sigma_param, -0.5)
    for m in actor_critic.modules():
        if isinstance(m, torch.nn.Linear):
            # orthogonal initialization
            torch.nn.init.orthogonal_(m.weight, gain=np.sqrt(2))
            torch.nn.init.zeros_(m.bias)
    # do last policy layer scaling, this will make initial actions have (close to)
    # 0 mean and std, and will help boost performances,
    # see https://arxiv.org/abs/2006.05990, Fig.24 for details
    for m in actor.mu.modules():
        if isinstance(m, torch.nn.Linear):
            torch.nn.init.zeros_(m.bias)
            m.weight.data.copy_(0.01 * m.weight.data)

    optim = torch.optim.Adam(actor_critic.parameters(), lr=args.lr)

    lr_scheduler = None
    if args.lr_decay:
        # decay learning rate to 0 linearly
        max_update_num = np.ceil(args.step_per_epoch / args.step_per_collect) * args.epoch

        lr_scheduler = LambdaLR(optim, lr_lambda=lambda epoch: 1 - epoch / max_update_num)

    def dist(loc_scale: tuple[torch.Tensor, torch.Tensor]) -> Distribution:
        loc, scale = loc_scale
        return Independent(Normal(loc, scale), 1)

    policy: PPOPolicy = PPOPolicy(
        actor=actor,
        critic=critic,
        optim=optim,
        dist_fn=dist,
        discount_factor=args.gamma,
        gae_lambda=args.gae_lambda,
        max_grad_norm=args.max_grad_norm,
        vf_coef=args.vf_coef,
        ent_coef=args.ent_coef,
        reward_normalization=args.rew_norm,
        action_scaling=True,
        action_bound_method=args.bound_action_method,
        lr_scheduler=lr_scheduler,
        action_space=env.action_space,
        eps_clip=args.eps_clip,
        value_clip=args.value_clip,
        dual_clip=args.dual_clip,
        advantage_normalization=args.norm_adv,
        recompute_advantage=args.recompute_adv,
    )

    # load a previous policy
    if args.resume_path:
        ckpt = torch.load(args.resume_path, map_location=args.device)
        policy.load_state_dict(ckpt["model"])
        train_envs.set_obs_rms(ckpt["obs_rms"])
        test_envs.set_obs_rms(ckpt["obs_rms"])
        print("Loaded agent from: ", args.resume_path)

    # collector
    buffer: VectorReplayBuffer | ReplayBuffer
    if args.training_num > 1:
        buffer = VectorReplayBuffer(args.buffer_size, len(train_envs))
    else:
        buffer = ReplayBuffer(args.buffer_size)
    train_collector = Collector[CollectStats](policy, train_envs, buffer, exploration_noise=True)
    test_collector = Collector[CollectStats](policy, test_envs)

    # log
    now = datetime.datetime.now().strftime("%y%m%d-%H%M%S")
    args.algo_name = "ppo"
    log_name = os.path.join(args.task, args.algo_name, str(args.seed), now)
    log_path = os.path.join(args.logdir, log_name)

    # logger
    logger_factory = LoggerFactoryDefault()
    if args.logger == "wandb":
        logger_factory.logger_type = "wandb"
        logger_factory.wandb_project = args.wandb_project
    else:
        logger_factory.logger_type = "tensorboard"

    logger = logger_factory.create_logger(
        log_dir=log_path,
        experiment_name=log_name,
        run_id=args.resume_id,
        config_dict=vars(args),
    )

    def save_best_fn(policy: BasePolicy) -> None:
        state = {"model": policy.state_dict(), "obs_rms": train_envs.get_obs_rms()}
        torch.save(state, os.path.join(log_path, "policy.pth"))

    if not args.watch:
        # trainer
        result = OnpolicyTrainer(
            policy=policy,
            train_collector=train_collector,
            test_collector=test_collector,
            max_epoch=args.epoch,
            step_per_epoch=args.step_per_epoch,
            repeat_per_collect=args.repeat_per_collect,
            episode_per_test=args.test_num,
            batch_size=args.batch_size,
            step_per_collect=args.step_per_collect,
            save_best_fn=save_best_fn,
            logger=logger,
            test_in_train=False,
        ).run()
        pprint.pprint(result)

    # Let's watch its performance!
    test_envs.seed(args.seed)
    test_collector.reset()
    collector_stats = test_collector.collect(n_episode=args.test_num, render=args.render)
    print(collector_stats)


if __name__ == "__main__":
    test_ppo()