def _make_env(config_file, host, port):
config = load_config(config_file)
if config['typename'] == 'RemoteEnv':
if port:
config['args']['port'] = port
if host:
config['args']['host'] = host
return get_env(config['typename'])(**config.get('args', {}))
def _build_network(model_filepath, optimizer_filepath, initial_parameter):
_LG.info('Building Q networks')
dql = DeepQLearning(
q_learning_config={
'discount_rate': 0.99,
'min_reward': -1.0,
'max_reward': 1.0,
},
optimizer_config=load_config(optimizer_filepath),
)
model_def = _gen_model_def(model_filepath)
dql.build(model_def, initial_parameter)
_LG.info('Syncing models')
dql.sync_network()
return dql
def get_model_config(filepath, **parameters):
"""Load model configurations from file
Parameters
----------
model_name : str
Model name or path to YAML file
parameters
Parameter for model config
Returns
-------
JSON-compatible object
Model configuration.
"""
if not os.path.isfile(filepath):
raise ValueError(
'Model definition file ({}) was not found.'.format(filepath))
return load_config(filepath, **parameters)
def _load_optimizer(filepath):
cfg = load_config(filepath)
return get_optimizer(cfg['typename'])(**cfg.get('args', {}))
def _make_agent(config_file):
config = (load_config(config_file) if config_file else {
'typename': 'NoOpAgent',
'args': {}
})
return get_agent(config['typename'])(**config.get('args', {}))
def create_env(cfg_file):
"""Load Environment config file and instantiate"""
cfg = load_config(cfg_file)
env = get_env(cfg['name'])(**cfg['args'])
print('\n{}'.format(env))
return env
def _make_optimizer(filepath):
cfg = load_config(filepath)
return nn.get_optimizer(cfg['typename'])(**cfg['args'])
def _load_agent(config_file_path):
config = load_config(config_file_path)
return get_agent(config['name'])(**config['args'])
def _load_optimizer(filepath):
cfg = load_config(filepath)
return nn.fetch_optimizer(cfg['typename'])(**cfg.get('args', {}))