def __init__(self):
self._alpha = self.ALPHA # Learning rate
self._gamma = self.GAMMA # Future reward discount
self._state_interpreter = StateInterpreter()
self._device = 'cuda' if torch.cuda.is_available() else 'cpu'
self._policy_net = self._create_neural_network()
self._loss = nn.MSELoss()
self._optim = optim.Adam(self._policy_net.parameters(), self._alpha)
if self.LOAD_PATH is not None:
self.load(self.LOAD_PATH)
self._save_dir_path = Utils.get_base_dir().joinpath(self.SAVE_DIR)
def _create_default_log(self) -> None:
log_dir_name = 'log'
project_path = Utils.get_base_dir()
log_dir_path = project_path.joinpath(log_dir_name)
Utils.create_directory(log_dir_path)
self._create_unique_log_file(log_dir_path)
def _generate_log_dir_path(self) -> str:
comment = self._generate_comment()
return str(Utils.get_base_dir().joinpath('runs').joinpath(
f'{Utils.get_now_as_str()}: {comment}'))
