Hands-On Neuroevolution with Python

Running work tasks and monitoring results

To use the RL evaluation mechanism we described earlier, we need a method to schedule the work task for evaluation and to monitor the results. This is implemented in the monitor_eval function, which receives the list of genomes in the population and evaluates them against the Atari game environment. This function has two essential implementation parts, both of which we'll discuss in this section:

1. First, we iterate over all the genomes in the list and create the asynchronous work task so that each genome can be evaluated against the Atari game environment:

    tasks = []    for t in it:        tasks.append(self.eval_async(*t, max_frames=max_frames,                                     error_callback=error_callback)) if time.time() - tstart > logging_interval: ...