def register_algorithms_and_preprocessors(self):
try:
from custom.algorithms import CUSTOM_ALGORITHMS
from custom.preprocessors import CUSTOM_PREPROCESSORS
from custom.models.impala_cnn_tf import ImpalaCNN
except ModuleNotFoundError:
from algorithms import CUSTOM_ALGORITHMS
from preprocessors import CUSTOM_PREPROCESSORS
from models.impala_cnn_tf import ImpalaCNN
load_algorithms(CUSTOM_ALGORITHMS)
load_preprocessors(CUSTOM_PREPROCESSORS)
ModelCatalog.register_custom_model("impala_cnn_tf", ImpalaCNN)
def register_algorithms_and_preprocessors(self):
# TODO(annaluo@): register custom model via dynamic sourcing
try:
from custom.algorithms import CUSTOM_ALGORITHMS
from custom.preprocessors import CUSTOM_PREPROCESSORS
from custom.models.my_vision_network import MyVisionNetwork
except ModuleNotFoundError:
from algorithms import CUSTOM_ALGORITHMS
from preprocessors import CUSTOM_PREPROCESSORS
from models.my_vision_network import MyVisionNetwork
load_algorithms(CUSTOM_ALGORITHMS)
load_preprocessors(CUSTOM_PREPROCESSORS)
ModelCatalog.register_custom_model("my_vision_network",
MyVisionNetwork)
Training with Config:
python ./train.py -f experiments/simple-corridor-0.yaml
Note that -f overrides all other trial-specific command-line options.
"""
# Register all necessary assets in tune registries
load_envs(os.getcwd()) # Load envs
load_models(os.getcwd()) # Load models
# Load custom algorithms
from algorithms import CUSTOM_ALGORITHMS
load_algorithms(CUSTOM_ALGORITHMS)
# Load custom preprocessors
from preprocessors import CUSTOM_PREPROCESSORS
load_preprocessors(CUSTOM_PREPROCESSORS)
print(ray.rllib.contrib.registry.CONTRIBUTED_ALGORITHMS)
def create_parser(parser_creator=None):
parser = make_parser(parser_creator=parser_creator,
formatter_class=argparse.RawDescriptionHelpFormatter,
description="Train a reinforcement learning agent.",
epilog=EXAMPLE_USAGE)
# See also the base parser definition in ray/tune/config_parser.py
parser.add_argument(
"--ray-address",
default=None,
type=str,
EXAMPLE_USAGE = """
Training example:
python ./train.py --run DQN --env CartPole-v0
Training with Config:
python ./train.py -f experiments/simple-corridor-0.yaml
Note that -f overrides all other trial-specific command-line options.
"""
# Register all necessary assets in tune registries
load_envs(os.getcwd()) # Load envs
load_models(os.getcwd()) # Load models
load_preprocessors(os.getcwd()) # Load preprocessors
def create_parser(parser_creator=None):
parser = make_parser(parser_creator=parser_creator,
formatter_class=argparse.RawDescriptionHelpFormatter,
description="Train a reinforcement learning agent.",
epilog=EXAMPLE_USAGE)
# See also the base parser definition in ray/tune/config_parser.py
parser.add_argument(
"--ray-address",
default=None,
type=str,
help="Connect to an existing Ray cluster at this address instead "
"of starting a new one.")