def test_should_be_able_to_set_property(self):
reinforce_config = ReinforceConfig()
self.assertEqual(reinforce_config.decay_steps, 250)
reinforce_config.decay_steps = 100
self.assertEqual(reinforce_config.decay_steps, 100)
def main():
parser = argparse.ArgumentParser()
parser.add_argument(
'-f', '--folder',
help='The folder containing the config files',
required=True
)
parser.add_argument(
'-m', '--map',
help='Run the specified map',
required=False
)
parser.add_argument(
'--eval',
help="Run only evaluation task",
dest='eval',
action="store_true"
)
parser.add_argument(
'-r', '--render',
help="Render task",
dest='render',
action="store_true"
)
args = parser.parse_args()
evaluation_config_path = os.path.join(args.folder, "evaluation.yml")
evaluation_config = EvaluationConfig.load_from_yaml(evaluation_config_path)
network_config_path = os.path.join(args.folder, "network.yml")
network_config = NetworkConfig.load_from_yaml(network_config_path)
reinforce_config_path = os.path.join(args.folder, "reinforce.yml")
reinforce_config = ReinforceConfig.load_from_yaml(reinforce_config_path)
map_name = args.map
if map_name is None:
print("You are traning the agent for the default map: ")
print("FourTowersWithFriendlyUnitsFixedEnemiesFixedPosition")
else:
print("You are traning the agent for the map: ")
print(map_name)
#print(map_name)
if args.eval:
evaluation_config.training_episodes = 0
network_config.restore_network = True
if args.render:
evaluation_config.render = True
run_task(evaluation_config, network_config, reinforce_config, map_name = map_name)
return 0
def test_should_have_default_values(self):
reinforce_config = ReinforceConfig()
self.assertEqual(reinforce_config.decay_steps, 250)
self.assertEqual(reinforce_config.starting_epsilon, 1.0)
self.assertEqual(reinforce_config.decay_rate, 0.96)
self.assertEqual(reinforce_config.discount_factor, 0.95)
self.assertEqual(reinforce_config.batch_size, 35)
self.assertEqual(reinforce_config.memory_size, 10000)
self.assertEqual(reinforce_config.summaries_path, None)
def test_should_create_reinforce_config_object_from_file(self):
reinforce_config = ReinforceConfig.load_from_yaml(test_file)
self.assertEqual(reinforce_config.decay_steps, 500)
self.assertEqual(reinforce_config.starting_epsilon, 0.9)
self.assertEqual(reinforce_config.decay_rate, 0.80)
self.assertEqual(reinforce_config.discount_factor, 1)
self.assertEqual(reinforce_config.batch_size, 10)
self.assertEqual(reinforce_config.memory_size, 20000)
self.assertEqual(reinforce_config.summaries_path,
"path/to/reinforcement/summaries.ckpt")
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-f',
'--folder',
help='The folder containing the config files',
required=True)
# TODO Better way to load the task to run
parser.add_argument(
'-t',
'--task',
help="The task to run. The python module cointaining the ABP program",
required=True)
parser.add_argument(
'-d',
'--debug',
help="Print lots of debugging statements",
action="store_const",
dest="loglevel",
const=logging.DEBUG,
default=logging.WARNING,
)
parser.add_argument(
'-v',
'--verbose',
help="Be verbose. Logging level INFO",
action="store_const",
dest="loglevel",
const=logging.INFO,
)
parser.add_argument('--eval',
help="Run only evaluation task",
dest='eval',
action="store_true")
parser.add_argument('-r',
'--render',
help="Render task",
dest='render',
action="store_true")
args = parser.parse_args()
logger.setLevel(args.loglevel)
evaluation_config_path = os.path.join(args.folder, "evaluation.yml")
evaluation_config = EvaluationConfig.load_from_yaml(evaluation_config_path)
network_config_path = os.path.join(args.folder, "network.yml")
network_config = NetworkConfig.load_from_yaml(network_config_path)
reinforce_config_path = os.path.join(args.folder, "reinforce.yml")
reinforce_config = ReinforceConfig.load_from_yaml(reinforce_config_path)
if args.eval:
evaluation_config.training_episodes = 0
network_config.restore_network = True
if args.render:
evaluation_config.render = True
task_module = import_module(args.task)
task_module.run_task(evaluation_config, network_config, reinforce_config)
return 0
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-f',
'--folder',
help='The folder containing the config files',
required=True)
parser.add_argument('-m',
'--map',
help='Run the specified map',
required=False)
parser.add_argument('-t',
'--test',
help='Just run test, no train',
dest='test',
action="store_true")
parser.add_argument(
'-tf',
'--train_forever',
help=
'After reach optimal policy. No need more eposdes, train the agent train_forever',
dest='train_forever',
action="store_true")
parser.add_argument('--eval',
help="Run only evaluation task",
dest='eval',
action="store_true")
parser.add_argument('-r',
'--render',
help="Render task",
dest='render',
action="store_true")
args = parser.parse_args()
#print(args)
#input()
evaluation_config_path = os.path.join(args.folder, "evaluation.yml")
evaluation_config = EvaluationConfig.load_from_yaml(evaluation_config_path)
network_config_path = os.path.join(args.folder, "network.yml")
network_config = NetworkConfig.load_from_yaml(network_config_path)
reinforce_config_path = os.path.join(args.folder, "reinforce.yml")
reinforce_config = ReinforceConfig.load_from_yaml(reinforce_config_path)
if args.test:
evaluation_config.training_episodes = 0
map_name = args.map
if map_name is None:
print("You are traning the agent for the default map: ")
print("FourTowesFriendlyunitsDecomposedGroupReward")
else:
print("You are traning the agent for the map: ")
print(map_name)
#print(map_name)
if args.eval:
evaluation_config.training_episodes = 0
network_config.restore_network = True
if args.render:
evaluation_config.render = True
run_task(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
return 0
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-f',
'--folder',
help='The folder containing the config files',
required=True)
parser.add_argument('-m',
'--map',
help='Run the specified map',
required=False)
parser.add_argument('-tk',
'--task',
help="which task to run",
dest='task',
required=False)
parser.add_argument('--model',
help="which model for agent",
dest='agent_model',
required=False)
parser.add_argument('-t',
'--test',
help='Just run test, no train',
dest='test',
action="store_true")
parser.add_argument('-ce',
'--collecting_experience',
help='Just run test and collect experience',
dest='collecting_experience',
action="store_true")
parser.add_argument(
'-tf',
'--train_forever',
help=
'After reach optimal policy. No need more eposdes, train the agent train_forever',
dest='train_forever',
action="store_true")
parser.add_argument('--eval',
help="Run only evaluation task",
dest='eval',
action="store_true")
parser.add_argument('-r',
'--render',
help="Render task",
dest='render',
action="store_true")
args = parser.parse_args()
evaluation_config_path = os.path.join(args.folder, "evaluation.yml")
evaluation_config = EvaluationConfig.load_from_yaml(evaluation_config_path)
network_config_path = os.path.join(args.folder, "network.yml")
network_config = NetworkConfig.load_from_yaml(network_config_path)
reinforce_config_path = os.path.join(args.folder, "reinforce.yml")
reinforce_config = ReinforceConfig.load_from_yaml(reinforce_config_path)
if args.test:
evaluation_config.training_episodes = 0
if args.collecting_experience:
reinforce_config.collecting_experience = True
map_name = args.map
if map_name is None:
print("You are traning the agent for the default map: ")
print("TugOfWar")
else:
print("You are traning the agent for the map: ")
print(map_name)
#print(map_name)
if args.eval:
evaluation_config.training_episodes = 0
network_config.restore_network = True
if args.render:
evaluation_config.render = True
if args.task == 'task_bigA':
run_task_bigA(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
elif args.task == 'task_single_player':
run_task(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
elif args.task == 'task_2p':
run_task_2p(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
elif args.task == 'task_mbts':
run_task_mbts(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
elif args.task == 'task_2p_2l':
run_task_2p_2l(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
elif args.task == 'task_2p_2l_grid':
run_task_2p_2l_grid(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever)
elif args.task == 'task_2p_2l_hp':
run_task_2p_2l_hp(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
train_forever=args.train_forever,
agent_model=args.agent_model)
elif args.task == 'task_2p_2l_deexplanation':
run_task_2p_2l_deexplanation(evaluation_config,
network_config,
reinforce_config,
map_name=map_name,
agent_model=args.agent_model)
elif args.task == 'task_2p_2l_grid_decomposed':
run_task_2p_2l_grid_decomposed(evaluation_config,
network_config,
reinforce_config,
map_name=map_name)
elif args.task == 'task_2p_2l_grid_decomposed_trans':
run_task_2p_2l_grid_decomposed_trans(evaluation_config,
network_config,
reinforce_config,
map_name=map_name)
elif args.task == 'task_mbts_grid':
run_task_mbts_grid(evaluation_config,
network_config,
reinforce_config,
map_name=map_name)
else:
print("need task")
return 0