def play_func(params, net, cuda, exp_queue):
env = atari_wrappers.make_atari(params.env_name,
skip_noop=True,
skip_maxskip=True)
env = atari_wrappers.wrap_deepmind(env,
pytorch_img=True,
frame_stack=True,
frame_stack_count=2)
env.seed(common.SEED)
device = torch.device("cuda" if cuda else "cpu")
selector = ptan.actions.EpsilonGreedyActionSelector(
epsilon=params.epsilon_start)
epsilon_tracker = common.EpsilonTracker(selector, params)
agent = ptan.agent.DQNAgent(net, selector, device=device)
exp_source = ptan.experience.ExperienceSourceFirstLast(env,
agent,
gamma=params.gamma)
for frame_idx, exp in enumerate(exp_source):
epsilon_tracker.frame(frame_idx / BATCH_MUL)
exp_queue.put(exp)
for reward, steps in exp_source.pop_rewards_steps():
exp_queue.put(EpisodeEnded(reward, steps, selector.epsilon))
"--params",
default="egreedy",
choices=list(HYPERPARAMS.keys()),
help="Parameters, default=egreedy",
)
args = parser.parse_args()
params = HYPERPARAMS[args.params]
device = torch.device("cuda" if args.cuda else "cpu")
envs = []
for _ in range(N_ENVS):
env = atari_wrappers.make_atari(params.env_name,
skip_noop=True,
skip_maxskip=True)
env = atari_wrappers.wrap_deepmind(env,
pytorch_img=True,
frame_stack=True)
envs.append(env)
epsilon_tracker = None
selector = ptan.actions.ArgmaxActionSelector()
if args.params == "egreedy":
net = dqn_extra.BaselineDQN(env.observation_space.shape,
env.action_space.n).to(device)
selector = ptan.actions.EpsilonGreedyActionSelector(
epsilon=params.epsilon_start)
epsilon_tracker = common.EpsilonTracker(selector, params)
elif args.params == "noisynet":
net = dqn_extra.NoisyDQN(env.observation_space.shape,
env.action_space.n).to(device)