def create_control(params: SimpleNamespace, config_name) -> LoopControlV10:
env = CarEnvV10(mode_energy_penalty = params.env_mode_energy_penalty,
mode_random = params.env_mode_random,
mode_limit_steps = params.env_mode_limit_steps,
mode_time_penalty = params.env_mode_time_penalty)
agent = SimpleAgentV10(env,
devicestr = params.agent_device,
gamma = params.agent_gamma_exp,
buffer_size = params.agent_buffer_size,
target_net_sync = params.agent_target_net_sync,
eps_start = params.agent_simple_eps_start,
eps_final = params.agent_simple_eps_final,
eps_frames = params.agent_simple_eps_frames,
)
bridge = SimpleBridgeV10(agent=agent,
optimizer = params.bridge_optimizer,
learning_rate = params.bridge_learning_rate,
gamma = params.bridge_gamma,
initial_population = params.bridge_initial_population,
batch_size = params.bridge_batch_size,
)
control = LoopControlV10(
bridge = bridge,
run_name = config_name,
bound_avg_reward = params.loop_bound_avg_reward,
logtb = params.loop_logtb,
logfolder = "./../runs/runv00")
return control
def create_control(params: SimpleNamespace, config_name) -> LoopControlV10:
universe = InvestUniverse()
env = RoboAdvisorEnvV10(universe,
reward_average_count = params.env_reward_average_count,
start_cash = params.env_start_cash,
trading_cost = params.env_trading_cost,
buy_volume = params.env_buy_volumne,
proportional_buy_volume = params.env_proportional_buy_volume)
agent = RoboAdvisorAgentV10(env,
devicestr = params.agent_device,
gamma = params.agent_gamma_exp,
buffer_size = params.agent_buffer_size,
target_net_sync = params.agent_target_net_sync,
eps_start = params.agent_simple_eps_start,
eps_final = params.agent_simple_eps_final,
eps_frames = params.agent_simple_eps_frames,
hidden_size = params.agent_hidden_size ,
hidden_layers = params.agent_hidden_layers ,
dueling_network = params.agent_dueling_network ,
steps_count = params.agent_steps_count ,
use_combined_replay_buffer = params.agent_use_combined_replay_buffer,
)
bridge = SimpleBridgeV10(agent=agent,
output_actions = len(universe.get_companies()),
output_action_states = 3,
optimizer = params.bridge_optimizer,
learning_rate = params.bridge_learning_rate,
gamma = params.bridge_gamma,
initial_population = params.bridge_initial_population,
batch_size = params.bridge_batch_size,
)
control = LoopControlV10(
bridge = bridge,
run_name = config_name,
bound_avg_reward = params.loop_bound_avg_reward,
logtb = params.loop_logtb,
logfolder = "./../runs/runv00")
return control
def basic_simple_init(devicestr="cpu") -> SimpleBridgeV10:
env = CarEnvV10()
agent = SimpleAgentV10(env, devicestr, gamma=0.9, buffer_size=1000)
bridge = SimpleBridgeV10(agent, gamma=0.9)
return bridge
def basic_simple_init(devicestr="cpu") -> SimpleBridgeV10:
env = RoboAdvisorEnvV10(universe)
agent = RoboAdvisorAgentV10(env, devicestr, gamma=0.9, buffer_size=1000)
bridge = SimpleBridgeV10(agent, gamma=0.9)
return bridge
def basic_init_bridge() -> SimpleBridgeV10:
env = CarEnvV10()
agent = SimpleAgentV10(env, "cpu", gamma=0.9, buffer_size=1000)
bridge = SimpleBridgeV10(agent, gamma=0.9)
return bridge