def __init__(self, config):
DDQN.__init__(self, config)
self.memory = Prioritised_Replay_Buffer(self.hyperparameters,
config.seed)
if config.resume:
self.load_resume(config.resume_path)
def __init__(self, config, agent_name_=agent_name):
DDQN.__init__(self, config, agent_name_=agent_name_)
self.q_network_local = self.create_NN(input_dim=self.state_size, output_dim=self.action_size + 1)
self.q_network_optimizer = optim.Adam(self.q_network_local.parameters(), lr=self.hyperparameters["learning_rate"], eps=1e-4)
self.q_network_target = self.create_NN(input_dim=self.state_size, output_dim=self.action_size + 1)
Base_Agent.copy_model_over(from_model=self.q_network_local, to_model=self.q_network_target)
self.wandb_watch(self.q_network_local, log_freq=self.config.wandb_model_log_freq)
def __init__(self, config):
DDQN.__init__(self, config)
model_path = self.config.model_path if self.config.model_path else 'Models'
self.q_network_local = self.create_NN(input_dim=self.state_size, output_dim=self.action_size + 1)
self.q_network_local_path = os.path.join(model_path, "{}_q_network_local.pt".format(self.agent_name))
if self.config.load_model: self.locally_load_policy()
self.q_network_optimizer = optim.Adam(self.q_network_local.parameters(), lr=self.hyperparameters["learning_rate"], eps=1e-4)
self.q_network_target = self.create_NN(input_dim=self.state_size, output_dim=self.action_size + 1)
Base_Agent.copy_model_over(from_model=self.q_network_local, to_model=self.q_network_target)
def __init__(self, config, agent_name_=agent_name):
DDQN.__init__(self, config, agent_name_=agent_name_)
self.memory = Prioritised_Replay_Buffer(self.hyperparameters,
config.seed)
def __init__(self, config):
DDQN.__init__(self, config)
self.memory = Prioritised_Replay_Buffer(self.hyperparameters,
config.seed, config.use_GPU)