def playGame(actor):
random_state = np.random.RandomState(123)
n_games = 10
seed = 123
seeds = generate_deterministic_seeds(seed, n_games)
env = gym.make("Pendulum-v0")
# To use
# env = gym.make("BipedalWalker-v3")
# agent = doDDPG()
rewards = []
for i_episode in trange(n_games):
np.random.seed(seeds[i_episode])
random.seed(seeds[i_episode])
env.seed(int(seeds[i_episode]))
state = env.reset()
done = False
total_reward = 0
cur_state = env.reset()
while not done:
env.render()
cur_state = cur_state.reshape((1, env.observation_space.shape[0]))
action = actor.act(cur_state)
# print(action)
new_state, reward, done, info = env.step(action)
total_reward += reward
if done:
print(env)
break
rewards.append(total_reward)
print("Avg reward: {:0.2f}".format(np.mean(rewards)))
print("Min reward: {:0.2f}".format(np.min(rewards)))
print("Max reward: {:0.2f}".format(np.max(rewards)))
print()
env.close()
parser.add_argument("-s", "--seed", help="Random seed.", type=np.uint32,
default=None)
parser.add_argument("-v", "--verbose", action="store_true")
parser.add_argument("-sh", "--show_histogram", action="store_true")
args = parser.parse_args()
return args
if __name__ == "__main__":
args = parse_args()
# Fix the random seed
if args.seed is not None:
seeds = generate_deterministic_seeds(args.seed, args.num_games)
else:
seeds = None
# Choose a Pacman agent
AgentClass = load_agent(args.agent)
agent = AgentClass()
from misio.pacman.pacman import LocalPacmanGameRunner
runner = LocalPacmanGameRunner(layout_dir=args.layout,
random_ghosts=args.random_ghosts,
show_window=not args.no_graphics,
zoom_window=args.zoom,
frame_time=args.frame_time)
games = []
for i in tqdm.trange(args.num_games, leave=False):
from tqdm import trange
import gym
import numpy as np
import random
import time
random_state = np.random.RandomState(123)
if __name__ == "__main__":
# env = gym.make('BipedalWalker-v2')
n_games = 100
seed = 123
seeds = generate_deterministic_seeds(seed, n_games)
env = gym.make("Pendulum-v0")
# env = gym.make("BipedalWalker-v2")
rewards = []
for i_episode in trange(n_games):
np.random.seed(seeds[i_episode])
random.seed(seeds[i_episode])
env.seed(int(seeds[i_episode]))
state = env.reset()
done = False
total_reward = 0
while not done:
# env.render()
# time.sleep(0.05)
