How to use the tianshou.data.Collector function in tianshou

# Let's watch its performance!
        env = gym.make(args.task)
        collector = Collector(policy, env)
        result = collector.collect(n_episode=1, render=args.render)
        print(f'Final reward: {result["rew"]}, length: {result["len"]}')
        collector.close()

    # here we define an imitation collector with a trivial policy
    if args.task == 'Pendulum-v0':
        env.spec.reward_threshold = -300  # lower the goal
    net = Actor(Net(1, args.state_shape, device=args.device),
                args.action_shape, args.max_action, args.device
                ).to(args.device)
    optim = torch.optim.Adam(net.parameters(), lr=args.il_lr)
    il_policy = ImitationPolicy(net, optim, mode='continuous')
    il_test_collector = Collector(il_policy, test_envs)
    train_collector.reset()
    result = offpolicy_trainer(
        il_policy, train_collector, il_test_collector, args.epoch,
        args.step_per_epoch // 5, args.collect_per_step, args.test_num,
        args.batch_size, stop_fn=stop_fn, save_fn=save_fn, writer=writer)
    assert stop_fn(result['best_reward'])
    train_collector.close()
    il_test_collector.close()
    if __name__ == '__main__':
        pprint.pprint(result)
        # Let's watch its performance!
        env = gym.make(args.task)
        collector = Collector(il_policy, env)
        result = collector.collect(n_episode=1, render=args.render)
        print(f'Final reward: {result["rew"]}, length: {result["len"]}')
        collector.close()

test_envs.seed(args.seed)
    # model
    net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(net, args.action_shape).to(args.device)
    critic = Critic(net).to(args.device)
    optim = torch.optim.Adam(list(
        actor.parameters()) + list(critic.parameters()), lr=args.lr)
    dist = torch.distributions.Categorical
    policy = A2CPolicy(
        actor, critic, optim, dist, args.gamma, gae_lambda=args.gae_lambda,
        vf_coef=args.vf_coef, ent_coef=args.ent_coef,
        max_grad_norm=args.max_grad_norm, reward_normalization=args.rew_norm)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # log
    log_path = os.path.join(args.logdir, args.task, 'a2c')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

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

    # trainer
    result = onpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.repeat_per_collect,
        args.test_num, args.batch_size, stop_fn=stop_fn, save_fn=save_fn,
        writer=writer)

env_fns = [lambda x=i: MyTestEnv(size=x, sleep=0) for i in [2, 3, 4, 5]]

    venv = SubprocVectorEnv(env_fns)
    dum = VectorEnv(env_fns)
    policy = MyPolicy()
    env = env_fns[0]()
    c0 = Collector(policy, env, ReplayBuffer(size=100, ignore_obs_next=False),
                   preprocess_fn)
    c0.collect(n_step=3, log_fn=logger.log)
    assert np.allclose(c0.buffer.obs[:3], [0, 1, 0])
    assert np.allclose(c0.buffer[:3].obs_next, [1, 2, 1])
    c0.collect(n_episode=3, log_fn=logger.log)
    assert np.allclose(c0.buffer.obs[:8], [0, 1, 0, 1, 0, 1, 0, 1])
    assert np.allclose(c0.buffer[:8].obs_next, [1, 2, 1, 2, 1, 2, 1, 2])
    c0.collect(n_step=3, random=True)
    c1 = Collector(policy, venv, ReplayBuffer(size=100, ignore_obs_next=False),
                   preprocess_fn)
    c1.collect(n_step=6)
    assert np.allclose(c1.buffer.obs[:11], [0, 1, 0, 1, 2, 0, 1, 0, 1, 2, 3])
    assert np.allclose(c1.buffer[:11].obs_next,
                       [1, 2, 1, 2, 3, 1, 2, 1, 2, 3, 4])
    c1.collect(n_episode=2)
    assert np.allclose(c1.buffer.obs[11:21], [0, 1, 2, 3, 4, 0, 1, 0, 1, 2])
    assert np.allclose(c1.buffer[11:21].obs_next,
                       [1, 2, 3, 4, 5, 1, 2, 1, 2, 3])
    c1.collect(n_episode=3, random=True)
    c2 = Collector(policy, dum, ReplayBuffer(size=100, ignore_obs_next=False),
                   preprocess_fn)
    c2.collect(n_episode=[1, 2, 2, 2])
    assert np.allclose(c2.buffer.obs_next[:26], [
        1, 2, 1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 4, 5,
        1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 4, 5])

optim = torch.optim.Adam(list(
        actor.parameters()) + list(critic.parameters()), lr=args.lr)
    dist = torch.distributions.Categorical
    policy = PPOPolicy(
        actor, critic, optim, dist, args.gamma,
        max_grad_norm=args.max_grad_norm,
        eps_clip=args.eps_clip,
        vf_coef=args.vf_coef,
        ent_coef=args.ent_coef,
        action_range=None,
        gae_lambda=args.gae_lambda,
        reward_normalization=args.rew_norm,
        dual_clip=args.dual_clip,
        value_clip=args.value_clip)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # log
    log_path = os.path.join(args.logdir, args.task, 'ppo')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

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

    # trainer
    result = onpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.repeat_per_collect,

# test_envs = gym.make(args.task)
    test_envs = SubprocVectorEnv(
        [lambda: gym.make(args.task) for _ in range(args.test_num)],
        reset_after_done=False)
    # seed
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    train_envs.seed(args.seed)
    test_envs.seed(args.seed)
    # model
    net = Net(args.layer_num, args.state_shape, args.action_shape, args.device)
    net = net.to(args.device)
    optim = torch.optim.Adam(net.parameters(), lr=args.lr)
    policy = DQNPolicy(net, optim, args.gamma, args.n_step)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs, stat_size=args.test_num)
    train_collector.collect(n_step=args.batch_size)
    # log
    writer = SummaryWriter(args.logdir)

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

    def train_fn(x):
        policy.sync_weight()
        policy.set_eps(args.eps_train)

    def test_fn(x):
        policy.set_eps(args.eps_test)

critic1 = Critic(net, args.device).to(args.device)
    critic1_optim = torch.optim.Adam(critic1.parameters(), lr=args.critic_lr)
    critic2 = Critic(net, args.device).to(args.device)
    critic2_optim = torch.optim.Adam(critic2.parameters(), lr=args.critic_lr)
    policy = TD3Policy(
        actor, actor_optim, critic1, critic1_optim, critic2, critic2_optim,
        args.tau, args.gamma, GaussianNoise(sigma=args.exploration_noise),
        args.policy_noise, args.update_actor_freq, args.noise_clip,
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=args.rew_norm,
        ignore_done=args.ignore_done,
        estimation_step=args.n_step)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # train_collector.collect(n_step=args.buffer_size)
    # log
    log_path = os.path.join(args.logdir, args.task, 'td3')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

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

    # trainer
    result = offpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.test_num,
        args.batch_size, stop_fn=stop_fn, save_fn=save_fn, writer=writer)

torch.manual_seed(args.seed)
    train_envs.seed(args.seed)
    test_envs.seed(args.seed)
    # model
    net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(net, args.action_shape).to(args.device)
    critic = Critic(net).to(args.device)
    optim = torch.optim.Adam(list(
        actor.parameters()) + list(critic.parameters()), lr=args.lr)
    dist = torch.distributions.Categorical
    policy = A2CPolicy(
        actor, critic, optim, dist, args.gamma, gae_lambda=args.gae_lambda,
        vf_coef=args.vf_coef, ent_coef=args.ent_coef,
        max_grad_norm=args.max_grad_norm, reward_normalization=args.rew_norm)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # log
    log_path = os.path.join(args.logdir, args.task, 'a2c')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

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

    # trainer
    result = onpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.repeat_per_collect,

actor = Actor(net, args.action_shape, args.max_action,
                  args.device).to(args.device)
    actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic = Critic(net, args.device).to(args.device)
    critic_optim = torch.optim.Adam(critic.parameters(), lr=args.critic_lr)
    policy = DDPGPolicy(
        actor, actor_optim, critic, critic_optim,
        args.tau, args.gamma, GaussianNoise(sigma=args.exploration_noise),
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=True, ignore_done=True)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # log
    writer = SummaryWriter(args.logdir + '/' + 'ddpg')

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

    # trainer
    result = offpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.test_num,
        args.batch_size, stop_fn=stop_fn, writer=writer)
    assert stop_fn(result['best_reward'])
    train_collector.close()
    test_collector.close()
    if __name__ == '__main__':
        pprint.pprint(result)

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

    # trainer
    result = offpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.test_num,
        args.batch_size, stop_fn=stop_fn, save_fn=save_fn, writer=writer)
    assert stop_fn(result['best_reward'])
    train_collector.close()
    test_collector.close()
    if __name__ == '__main__':
        pprint.pprint(result)
        # Let's watch its performance!
        env = gym.make(args.task)
        collector = Collector(policy, env)
        result = collector.collect(n_episode=1, render=args.render)
        print(f'Final reward: {result["rew"]}, length: {result["len"]}')
        collector.close()

actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic1 = Critic(net, args.device).to(args.device)
    critic1_optim = torch.optim.Adam(critic1.parameters(), lr=args.critic_lr)
    critic2 = Critic(net, args.device).to(args.device)
    critic2_optim = torch.optim.Adam(critic2.parameters(), lr=args.critic_lr)
    policy = SACPolicy(
        actor, actor_optim, critic1, critic1_optim, critic2, critic2_optim,
        args.tau, args.gamma, args.alpha,
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=True, ignore_done=True)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # train_collector.collect(n_step=args.buffer_size)
    # log
    log_path = os.path.join(args.logdir, args.task, 'sac', args.run_id)
    writer = SummaryWriter(log_path)

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

    # trainer
    result = offpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.test_num,
        args.batch_size, stop_fn=stop_fn,
        writer=writer, log_interval=args.log_interval)
    assert stop_fn(result['best_reward'])
    train_collector.close()