def cli():
parser = argparse.ArgumentParser(
description="Train or evaluate an RLlib Trainer.",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=EXAMPLE_USAGE)
subcommand_group = parser.add_subparsers(
help="Commands to train or evaluate an RLlib agent.", dest="command")
# see _SubParsersAction.add_parser in
# https://github.com/python/cpython/blob/master/Lib/argparse.py
train_parser = train.create_parser(
lambda **kwargs: subcommand_group.add_parser("train", **kwargs))
evaluate_parser = evaluate.create_parser(
lambda **kwargs: subcommand_group.add_parser("evaluate", **kwargs))
rollout_parser = evaluate.create_parser(
lambda **kwargs: subcommand_group.add_parser("rollout", **kwargs))
options = parser.parse_args()
if options.command == "train":
train.run(options, train_parser)
elif options.command == "evaluate":
evaluate.run(options, evaluate_parser)
elif options.command == "rollout":
deprecation_warning(old="rllib rollout",
new="rllib evaluate",
error=False)
evaluate.run(options, rollout_parser)
else:
parser.print_help()
def train_agent(config):
parser = argparse.ArgumentParser(
description="Train or Run an RLlib Agent.",
formatter_class=argparse.RawDescriptionHelpFormatter)
subcommand_group = parser.add_subparsers(
help="Commands to train or run an RLlib agent.", dest="command")
train_parser = train.create_parser(
lambda **kwargs: subcommand_group.add_parser("train", **kwargs))
rollout_parser = rollout.create_parser(
lambda **kwargs: subcommand_group.add_parser("rollout", **kwargs))
options = parser.parse_args()
register_env('env_mp500lwa4dpg70', env_creator)
configFromYaml = {
'train-ppo': {
'env': 'env_mp500lwa4dpg70',
'run': 'PPO',
'config': config,
'checkpoint_freq': 1000,
#'local_dir': "~/train_results",
'stop': {
"timesteps_total": 1e10,
},
}
}
ray.init()
if options.command == "train":
configFromYaml['train-ppo']['config']['env_config'] = env_input_config(
True)
run_experiments(configFromYaml)
#train.run(options, train_parser)
elif options.command == "rollout":
options.run = configFromYaml['train-ppo']['run']
options.env = configFromYaml['train-ppo']['env']
options.no_render = True
options.steps = 1000
options.out = None
configFromYaml['train-ppo']['config']['env_config'] = env_input_config(
False)
configFromYaml['train-ppo']['config']['monitor'] = True
options.config = configFromYaml['train-ppo']['config']
rollout.run(options, rollout_parser)
else:
parser.print_help()
def cli():
parser = argparse.ArgumentParser(
description="Train or Run an RLlib Agent.",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=EXAMPLE_USAGE)
subcommand_group = parser.add_subparsers(
help="Commands to train or run an RLlib agent.", dest="command")
# see _SubParsersAction.add_parser in
# https://github.com/python/cpython/blob/master/Lib/argparse.py
train_parser = train.create_parser(
lambda **kwargs: subcommand_group.add_parser("train", **kwargs))
rollout_parser = rollout.create_parser(
lambda **kwargs: subcommand_group.add_parser("rollout", **kwargs))
options = parser.parse_args()
if options.command == "train":
train.run(options, train_parser)
elif options.command == "rollout":
rollout.run(options, rollout_parser)
else:
parser.print_help()
import ray
from ray.rllib import train
from ray import tune
import os
from utils import callbacks
if __name__ == "__main__":
# Parse arguments and add callbacks to config
train_parser = train.create_parser()
args = train_parser.parse_args()
args.config["callbacks"] = {"on_train_result": callbacks.on_train_result}
# Trace if video capturing is on
if 'monitor' in args.config and args.config['monitor']:
print("Video capturing is ON!")
# Start ray head (single node)
os.system('ray start --head')
ray.init(address='auto')
# Run training task using tune.run
tune.run(run_or_experiment=args.run,
config=dict(args.config, env=args.env),
stop=args.stop,
checkpoint_freq=args.checkpoint_freq,
checkpoint_at_end=args.checkpoint_at_end,
local_dir=args.local_dir)
activation_fn=None,
weights_initializer=weights_initializer)
return net_out, net
def create_pacman_environment(layout_name='originalClassic',
stick_actions=False):
"""
"""
layout = pacman_env.layout.getLayout(layout_name)
if layout is None:
raise ValueError('No suck layout as %s' % layout_name)
ghosts = []
for i in range(2):
ghosts.append(pacman_env.ghostAgents.RandomGhost(i + 1))
#display = VizGraphics(includeInfoPane=False, zoom=0.4)
display = pacman_env.matrixDisplay.PacmanGraphics(layout)
#teacherAgents = [LeftTurnAgent(), GreedyAgent()]
env = pacman_env.PacmanEnv(layout, ghosts,
display) #, teacherAgents=teacherAgents)
return env
if __name__ == '__main__':
ModelCatalog.register_custom_model("PacmanModel", PacmanModel)
register_env("pacman", lambda _: create_pacman_environment())
parser = create_parser()
args = parser.parse_args()
run(args, parser)
def main():
parser = create_parser()
args = parser.parse_args()
run(args, parser)
def create_parser():
parser = train.create_parser()
parser.set_defaults(env='CavalryVsInfantry')
return parser
def main():
raylab.register_all_agents()
raylab.register_all_environments()
parser = create_parser()
args = parser.parse_args()
run(args, parser)
