def create_env(cls, flags, featurizer=None):
f = featurizer or X.Concat([X.Text(), X.ValidMoves()])
print('loading env')
start_time = time.time()
env = gym.make(flags.env, room_shape=(flags.height, flags.width), partially_observable=flags.partial_observability, max_placement=flags.max_placement, featurizer=f, shuffle_wiki=flags.shuffle_wiki, time_penalty=flags.time_penalty)
print('loaded env in {} seconds'.format(time.time() - start_time))
return env
def create_env(cls, flags, featurizer=None):
return super().create_env(flags, featurizer=featurizer or X.Concat([X.Text(), X.ValidMoves(), X.RelativePosition()]))
from rtfm.dynamics.monster import Player
from rtfm import featurizer as F
if __name__ == '__main__':
parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--env', help='environment to use', default='rock_paper_scissors')
parser.add_argument('--height', type=int, help='room size', default=12)
parser.add_argument('--width', type=int, help='room size', default=22)
parser.add_argument('--seed', type=int, help='seed', default=42)
parser.add_argument('-p', '--partially_observable', action='store_true', help='only show partial observability')
parser.add_argument('-c', '--control', action='store_true', help='assume direct control')
parser.add_argument('-w', '--shuffle_wiki', action='store_true', help='shuffle facts in the wiki')
args = parser.parse_args()
featurizer = F.Concat([F.Progress(), F.ValidMoves(), F.Terminal()])
env = gym.make('rtfm:{}-v0'.format(args.env), featurizer=featurizer, partially_observable=args.partially_observable, room_shape=(args.height, args.width), shuffle_wiki=args.shuffle_wiki)
random.seed(args.seed)
score = total = 0
while True:
env.reset()
feat = reward = finished = won = None
while not finished:
if args.control:
ch = None
while ch not in Player.keymap:
print('Current score {} out of {}'.format(score, total))
print('Enter your command. x to quit.')
ch = getch.getch()