3735a6f690676bf2ab29f57ca1a02c1c63565b6c,ch16/02_breakout_es.py,,,#,179

Before Change

        writer.add_scalar("batch_steps", batch_steps, step_idx)
        speed = batch_steps / (time.time() - t_start)
        writer.add_scalar("speed", speed, step_idx)
        print("%d: reward=%.2f, speed=%.2f f/s" % (step_idx, m_reward, speed))

After Change

                break
            time.sleep(0.1)

        dt_data = time.time() - t_start
        m_reward = np.mean(batch_reward)
        if m_reward > 18:
            print("Solved in %d steps" % step_idx)
            break

        train_step(net, batch_noise, batch_reward, writer, step_idx)
        writer.add_scalar("reward_mean", m_reward, step_idx)
        writer.add_scalar("reward_std", np.std(batch_reward), step_idx)
        writer.add_scalar("reward_max", np.max(batch_reward), step_idx)
        writer.add_scalar("batch_episodes", len(batch_reward), step_idx)
        writer.add_scalar("batch_steps", batch_steps, step_idx)
        speed = batch_steps / (time.time() - t_start)
        writer.add_scalar("speed", speed, step_idx)
        dt_step = time.time() - t_start - dt_data

        print("%d: reward=%.2f, speed=%.2f f/s, data_gather=%.3f, train=%.3f, steps_mean=%.2f, min=%.2f, max=%.2f, steps_std=%.2f" % (
            step_idx, m_reward, speed, dt_data, dt_step, np.mean(batch_steps_data),
            np.min(batch_steps_data), np.max(batch_steps_data), np.std(batch_steps_data)))

Instances