def __init__(self, env, worker_id, gamma, env_render, logger, verbose):
self.env = env
self.worker_id = worker_id
# discount rate
self.gamma = gamma
self.trajectory = []
# learning rate
self.learning_rate = LEARNING_RATE
self.env_render = env_render
self.logger = logger
self.verbose = verbose
self.policy_model = rl_utils.get_rl_model(self.env).to(device)
self.target_model = rl_utils.get_rl_model(self.env).to(device)
self.target_model.load_state_dict(self.policy_model.state_dict())
self.target_model.eval()
self.optimizer = rl_utils.get_optimizer(
parameters=self.policy_model.parameters(),
learning_rate=self.learning_rate)
self.memory = ReplayMemory(10000)
self.steps_done = 0
self.model = self.policy_model
def __init__(self, env, worker_id, gamma, env_render, logger, verbose):
self.env = env
self.worker_id = worker_id
# discount rate
self.gamma = gamma
self.trajectory = []
# learning rate
self.learning_rate = LEARNING_RATE
self.env_render = env_render
self.logger = logger
self.verbose = verbose
self.model = rl_utils.get_rl_model(self.env, self.worker_id)
self.optimizer = rl_utils.get_optimizer(
parameters=self.model.parameters(),
learning_rate=self.learning_rate
)
os.environ["CUDA_VISIBLE_DEVICES"] = CUDA_VISIBLE_DEVICES_NUMBER_LIST
if __name__ == "__main__":
torch.manual_seed(SEED)
if torch.cuda.is_available():
torch.cuda.manual_seed(SEED)
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True
utils.make_output_folders()
utils.ask_file_removal()
env = rl_utils.get_environment()
rl_model = rl_utils.get_rl_model(env, -1)
utils.print_configuration(env, rl_model)
try:
chief = Process(target=utils.run_chief, args=())
chief.start()
time.sleep(1.5)
workers = []
for worker_id in range(NUM_WORKERS):
worker = Process(target=utils.run_worker, args=(worker_id, ))
workers.append(worker)
worker.start()