def play():
_vzd_path = os.path.dirname(vizdoom.__file__)
parser = argparse.ArgumentParser(description='Doom Network')
parser.add_argument('--action_set',
default='noset',
help='model to work with')
parser.add_argument('--doom_instance',
default='map',
choices=('basic', 'cig', 'map'),
help='doom instance type')
parser.add_argument('--vizdoom_config',
default='../environments/cig_test.cfg',
help='vizdoom config path')
parser.add_argument('--vizdoom_path',
default=_vzd_path,
help='path to vizdoom')
parser.add_argument('--wad_path',
default=_vzd_path + '/freedoom2.wad',
help='wad file path')
parser.add_argument('--skiprate',
type=int,
default=1,
help='number of skipped frames')
parser.add_argument('--frame_num',
type=int,
default=1,
help='number of frames per input')
parser.add_argument('--checkpoint_file',
default=None,
help='check point file name')
parser.add_argument('--checkpoint_rate',
type=int,
default=500,
help='number of batches per checkpoit')
parser.add_argument('--bot_cmd',
default=None,
help='command to launch a bot')
parser.add_argument('--h5_path', default=None, help='hd5 files path')
args = parser.parse_args()
print(args)
init_doom_env(args)
game = args.instance_class(args.vizdoom_config,
args.wad_path,
args.skiprate,
visible=True,
mode=vizdoom.Mode.PLAYER,
actions=args.action_set)
step_state = game.get_state_normalized()
for i in range(50):
step_state, _, finished = game.step_normalized([0, 0, 0, 0])
pose = game.get_pose()
state = game.get_state_normalized()
objects = game.get_objects(state)
sim = Simulator(pose, objects)
k = 1
default=_vzd_path,
help='path to vizdoom')
parser.add_argument('--wad_path',
default=_vzd_path + '/doom2.wad',
help='wad file path')
parser.add_argument('--skiprate',
type=int,
default=1,
help='number of skipped frames')
parser.add_argument('--frame_num',
type=int,
default=4,
help='number of frames per input')
args = parser.parse_args()
print(args)
init_doom_env(args)
torch.set_default_tensor_type('torch.FloatTensor')
torch.manual_seed(args.seed)
if USE_CUDA:
torch.cuda.manual_seed_all(args.seed)
model_class = {
'aac': AdvantageActorCritic,
'aac_lstm': AdvantageActorCriticLSTM,
'aac_intrinsic': AdvantageActorCriticIntrinsic,
'aac_duel': AdvantageActorCriticDuel
}
model = model_class[args.model](args)
if args.load is not None and os.path.isfile(args.load):
parser.add_argument('--epoch_game_steps', type=int, default=10000, help='number of steps per epoch')
parser.add_argument('--episode_discount', type=float, default=0.95, help='number of episodes for training')
parser.add_argument('--seed', type=int, default=1, help='seed value')
parser.add_argument(
'--model',
default='aac',
choices=('aac', 'aac_lstm', 'aac_noisy', 'aac_depth', 'aac_map', 'mb_map'),
help='model to work with')
parser.add_argument('--base_model', default=None, help='path to base model file')
parser.add_argument('--action_set', default=None, help='model to work with')
parser.add_argument('--load', default=None, help='path to model file')
parser.add_argument('--vizdoom_config', default='environments/basic.cfg', help='vizdoom config path')
parser.add_argument('--vizdoom_path', default=_vzd_path, help='path to vizdoom')
parser.add_argument('--wad_path', default=_vzd_path + '/freedoom2.wad', help='wad file path')
parser.add_argument('--skiprate', type=int, default=1, help='number of skipped frames')
parser.add_argument('--frame_num', type=int, default=1, help='number of frames per input')
parser.add_argument('--checkpoint_file', default=None, help='check point file name')
parser.add_argument('--checkpoint_rate', type=int, default=500, help='number of batches per checkpoit')
parser.add_argument('--bot_cmd', default=None, help='command to launch a bot')
parser.add_argument('--h5_path', default=None, help='hd5 files path')
args = parser.parse_args()
print(args)
init_doom_env(args)
model = get_model(args)
if args.mode == 'train':
model.run_train(args)
else:
model.run_test(args)
def main(_config):
""" Main script which for running a non-stationary doom environments
composed of multiple maps. """
# Setting logger
log = ex.logger
log.setLevel(_config['log_level'].upper())
# Printing the conf for visual check
log.debug(
'Exp. configurations:\n'
'----------------------------------------------------------------\n'
+ pprint.pformat(_config) + '\n' +
'----------------------------------------------------------------\n'
)
init_doom_env(_config)
random.seed(_config['seed'])
np.random.seed(_config['seed'])
model = get_model(_config)
if _config['mode'] == 'train':
test_crew = [[] for j in range(_config['num_runs'])]
test_std = [[] for j in range(_config['num_runs'])]
runs_avg_crew = [[[] for i in range(len(_config['vizdoom_maps']))]
for j in range(len(_config['vizdoom_maps']))]
# train with multiple runs
for run_id in range(_config['num_runs']):
ex.info[str(run_id)] = {'test_crew': [], 'test_std': []}
print("\n------------------ RUN {0} -------------------"
.format(run_id))
for task_id in range(len(_config['vizdoom_maps'])):
crew, std, movavg_stats = model.run_train(task_id=task_id)
test_crew[run_id].append(crew[-1])
test_std[run_id].append(std[-1])
# save results into mongodb
ex.info[str(run_id)]['test_crew'].append(crew[-1])
ex.info[str(run_id)]['test_std'].append(std[-1])
ex.info[str(run_id)]['mov_avg_stats'] = movavg_stats
ex.info[str(run_id)]['tot_crew'] = crew
ex.info[str(run_id)]['tot_std'] = std
# prepare for next run
del model
model = get_model(_config)
# printing results
print(
'----------------------------------------------------------------\n'
' RESULTS\n' +
'----------------------------------------------------------------\n'
)
for run_id in range(_config['num_runs']):
print("\n------------------ RUN {0} -------------------"
.format(run_id))
for task_id in range(len(_config['vizdoom_maps'])):
avg_crew = 0
avg_std = 0
for map_id in range(len(_config['vizdoom_maps'])):
# print(test_crew[run_id][task_id])
# print(test_std[run_id][task_id])
print(
"[run %d], [task: %d], [map: %d], "
"[avg. cumulated reward: %f.2f], "
"[dev.std: %.2f]" % (run_id, task_id, map_id,
test_crew[run_id][task_id][map_id],
test_std[run_id][task_id][map_id])
)
avg_crew += test_crew[run_id][task_id][map_id]
avg_std += test_std[run_id][task_id][map_id]
runs_avg_crew[task_id][map_id].append(
test_crew[run_id][task_id][map_id]
)
avg_crew /= len(_config['vizdoom_maps'])
avg_std /= len(_config['vizdoom_maps'])
print("[run %d] [avg_maps_crew %.3f], [avg_maps_std %.3f]\n" %
(run_id, avg_crew, avg_std))
print(
'----------------------------------------------------------------\n'
' AVG RUNS RESULTS\n' +
'----------------------------------------------------------------\n'
)
for task_id in range(len(_config['vizdoom_maps'])):
tot_avg_crew = []
for map_id in range(len(_config['vizdoom_maps'])):
print("[task %d], [map: %d], [runs avg_crew %.3f],"
" [runs std %.3f]" %
(task_id, map_id,
float(np.mean(runs_avg_crew[task_id][map_id])),
float(np.std(runs_avg_crew[task_id][map_id]))))
tot_avg_crew.append(np.mean(runs_avg_crew[task_id][map_id]))
print("[avg_runs_maps_crew %.3f], [std %.3f]\n" %
(float(np.mean(tot_avg_crew)), float(np.std(tot_avg_crew))))
else:
# we assume to have a list in load for test
res = {}
for model_path in _config['load']:
res[model_path] = []
for task_id in range(len(_config['vizdoom_maps'])):
crew, std = model.run_test(
num_test_ep=_config['num_test_ep'], load=model_path,
seed=_config['seed'], task_id=task_id)
res[model_path].append((crew, std))
res = OrderedDict(sorted(res.items(), key=lambda t: t[0]))
for model_name, vec in res.items():
avg_crew = 0
avg_std = 0
for i, (crew, std) in enumerate(vec):
print(
"[name %s], [map: %d], [avg. cumulated reward: %f.2f],"
"[dev.std: %.2f]" % (model_name, i, crew, std)
)
avg_crew += crew
avg_std += std
avg_crew /= len(res[model_name])
avg_std /= len(res[model_name])
print("[avg_crew %.3f], [avg_std %.3f]\n" % (avg_crew, avg_std))