def __init__(self, training_directory, config):
self.config = config
self.training_directory = training_directory
self.model_save_directory = self.training_directory + 'params/'
self.env = resolve_env(self.config['environment'])(test_cases[self.config['environment']][self.config['environment_index']](), self.training_directory, self.config)
self.env.pre_processing()
self.model = resolve_model(self.config['model'])(self.config)
self.MOR_flag = self.config['MOR_flag']
if (self.MOR_flag):
self.multiple_rewards = resolve_multiple_rewards(self.config['multiple_rewards'])
self.reward_mins = np.zeros(len(self.multiple_rewards))
self.reward_maxs = np.zeros(len(self.multiple_rewards))
self.master_params = self.model.init_master_params(self.config['from_file'], self.config['params_file'])
self.mu = self.config['n_individuals']/4
self.learning_rate = self.config['learning_rate'] * 20
self.A = np.sqrt(self.config['noise_std_dev']) * np.eye(len(self.master_params)) #sqrt of cov matrix
self.mu = self.config['mu']
for i in range(0, len(self.master_params)):
for j in range(i, len(self.master_params)):
self.A[i][j] += np.random.normal() * np.sqrt(self.config['noise_std_dev']) * 0.05
if (self.config['from_file']):
logging.info("\nLoaded Master Params from:")
logging.info(self.config['params_file'])
logging.info("\nReward:")
logging.info(inspect.getsource(self.reward) + "\n")
def __init__(self, training_directory, config):
self.config = config
self.training_directory = training_directory
self.model_save_directory = self.training_directory + 'params/'
self.env = resolve_env(self.config['environment'])(test_cases[self.config['environment']][self.config['environment_index']](), self.training_directory, self.config)
self.env.pre_processing()
self.model = resolve_model(self.config['model'])(self.config)
self.reward = resolve_reward(self.config['reward'])
self.MOR_flag = self.config['MOR_flag']
if (self.MOR_flag):
self.multiple_rewards = resolve_multiple_rewards(self.config['multiple_rewards'])
self.reward_mins = np.zeros(len(self.multiple_rewards))
self.reward_maxs = np.zeros(len(self.multiple_rewards))
self.master_params = self.model.init_master_params(self.config['from_file'], self.config['params_file'])
self.mu = self.config['n_individuals']/4
self.learning_rate = self.config['learning_rate']
self.noise_std_dev = self.config['noise_std_dev']
self.visualize = self.config['visualize']
self.visualize_every = self.config['visualize_every']
self.moving_success_rate = 0
if (self.config['from_file']):
logging.info("\nLoaded Master Params from:")
logging.info(self.config['params_file'])
logging.info("\nReward:")
logging.info(inspect.getsource(self.reward) + "\n")
def __init__(self, training_directory, config):
self.config = config
self.training_directory = training_directory
self.env = resolve_env(
self.config['environment'])(test_cases[self.config['environment']][
self.config['environment_index']], self.training_directory,
self.config)
self.env.pre_processing()
self.model = resolve_model(self.config['model'])(self.config)
self.reward = resolve_reward(self.config['reward'])
self.master_params = self.model.init_master_params()
self.learning_rate = self.config['learning_rate']
self.noise_std_dev = self.config['noise_std_dev']
logging.info("\nReward:")
logging.info(inspect.getsource(self.reward) + "\n")