1eb6137645
This is the PR for QR-DQN algorithm: https://arxiv.org/abs/1710.10044 1. add QR-DQN policy in tianshou/policy/modelfree/qrdqn.py. 2. add QR-DQN net in examples/atari/atari_network.py. 3. add QR-DQN atari example in examples/atari/atari_qrdqn.py. 4. add QR-DQN statement in tianshou/policy/init.py. 5. add QR-DQN unit test in test/discrete/test_qrdqn.py. 6. add QR-DQN atari results in examples/atari/results/qrdqn/. 7. add compute_q_value in DQNPolicy and C51Policy for simplify forward function. 8. move `with torch.no_grad():` from `_target_q` to BasePolicy By running "python3 atari_qrdqn.py --task "PongNoFrameskip-v4" --batch-size 64", get best_result': '19.8 ± 0.40', in epoch 8.
137 lines
5.1 KiB
Python
137 lines
5.1 KiB
Python
import os
|
|
import gym
|
|
import torch
|
|
import pprint
|
|
import argparse
|
|
import numpy as np
|
|
from torch.utils.tensorboard import SummaryWriter
|
|
|
|
from tianshou.policy import QRDQNPolicy
|
|
from tianshou.env import DummyVectorEnv
|
|
from tianshou.utils.net.common import Net
|
|
from tianshou.trainer import offpolicy_trainer
|
|
from tianshou.data import Collector, ReplayBuffer, PrioritizedReplayBuffer
|
|
|
|
|
|
def get_args():
|
|
parser = argparse.ArgumentParser()
|
|
parser.add_argument('--task', type=str, default='CartPole-v0')
|
|
parser.add_argument('--seed', type=int, default=1626)
|
|
parser.add_argument('--eps-test', type=float, default=0.05)
|
|
parser.add_argument('--eps-train', type=float, default=0.1)
|
|
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=0.9)
|
|
parser.add_argument('--num-quantiles', type=int, default=200)
|
|
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=1000)
|
|
parser.add_argument('--collect-per-step', type=int, default=10)
|
|
parser.add_argument('--batch-size', type=int, default=64)
|
|
parser.add_argument('--hidden-sizes', type=int,
|
|
nargs='*', default=[128, 128, 128, 128])
|
|
parser.add_argument('--training-num', type=int, default=8)
|
|
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.)
|
|
parser.add_argument('--prioritized-replay',
|
|
action="store_true", default=False)
|
|
parser.add_argument('--alpha', type=float, default=0.6)
|
|
parser.add_argument('--beta', type=float, default=0.4)
|
|
parser.add_argument(
|
|
'--device', type=str,
|
|
default='cuda' if torch.cuda.is_available() else 'cpu')
|
|
args = parser.parse_known_args()[0]
|
|
return args
|
|
|
|
|
|
def test_qrdqn(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
|
|
net = Net(args.state_shape, args.action_shape,
|
|
hidden_sizes=args.hidden_sizes, device=args.device,
|
|
softmax=False, num_atoms=args.num_quantiles)
|
|
optim = torch.optim.Adam(net.parameters(), lr=args.lr)
|
|
policy = QRDQNPolicy(
|
|
net, optim, args.gamma, args.num_quantiles,
|
|
args.n_step, target_update_freq=args.target_update_freq
|
|
).to(args.device)
|
|
# buffer
|
|
if args.prioritized_replay:
|
|
buf = PrioritizedReplayBuffer(
|
|
args.buffer_size, alpha=args.alpha, beta=args.beta)
|
|
else:
|
|
buf = ReplayBuffer(args.buffer_size)
|
|
# collector
|
|
train_collector = Collector(policy, train_envs, buf)
|
|
test_collector = Collector(policy, test_envs)
|
|
# policy.set_eps(1)
|
|
train_collector.collect(n_step=args.batch_size)
|
|
# log
|
|
log_path = os.path.join(args.logdir, args.task, 'qrdqn')
|
|
writer = SummaryWriter(log_path)
|
|
|
|
def save_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):
|
|
# eps annnealing, just a demo
|
|
if env_step <= 10000:
|
|
policy.set_eps(args.eps_train)
|
|
elif env_step <= 50000:
|
|
eps = args.eps_train - (env_step - 10000) / \
|
|
40000 * (0.9 * args.eps_train)
|
|
policy.set_eps(eps)
|
|
else:
|
|
policy.set_eps(0.1 * args.eps_train)
|
|
|
|
def test_fn(epoch, env_step):
|
|
policy.set_eps(args.eps_test)
|
|
|
|
# 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, train_fn=train_fn, test_fn=test_fn,
|
|
stop_fn=stop_fn, save_fn=save_fn, writer=writer)
|
|
|
|
assert stop_fn(result['best_reward'])
|
|
if __name__ == '__main__':
|
|
pprint.pprint(result)
|
|
# Let's watch its performance!
|
|
env = gym.make(args.task)
|
|
policy.eval()
|
|
policy.set_eps(args.eps_test)
|
|
collector = Collector(policy, env)
|
|
result = collector.collect(n_episode=1, render=args.render)
|
|
print(f'Final reward: {result["rew"]}, length: {result["len"]}')
|
|
|
|
|
|
def test_pqrdqn(args=get_args()):
|
|
args.prioritized_replay = True
|
|
args.gamma = .95
|
|
args.seed = 1
|
|
test_qrdqn(args)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
test_pqrdqn(get_args())
|