def run(self):
for x in range(self.config.total_step):
batch = self.replay_memory.sample()
if batch is not None:
training_info = self.model.step(batch)
if training_info is not None:
self.replay_memory.training_info(batch, training_info)
self.agent.update_actor(self.model.get_actor())
if x % 5000 == 0:
output_file_path = GLOBAL_LOGGER.get_log_path()
self.model.save(output_file_path, str(x))
GLOBAL_LOGGER.reset_event_file()
output_file_path = GLOBAL_LOGGER.get_log_path()
self.model.save(output_file_path, 'final')
def run(self):
for x in range(self.config.total_step):
t1 = time.time()
batch = self.replay_memory.sample(asynchronization=True)
if batch is not None:
training_info = self.model.step(batch)
if training_info is not None:
self.replay_memory.training_info(batch, training_info)
self.agent.update_actor(self.model.get_actor())
GLOBAL_LOGGER.get_tb_logger().add_scalar(
"training_time",
time.time() - t1, x)
else:
time.sleep(0.01)
if x % 5000 == 0:
output_file_path = GLOBAL_LOGGER.get_log_path()
self.model.save(output_file_path, str(x))
GLOBAL_LOGGER.reset_event_file()
output_file_path = GLOBAL_LOGGER.get_log_path()
self.model.save(output_file_path, 'final')