3 ms·It’s unclear to me how this could work with current training methods.by layer8 1y agoIt’s unclear to me how this could work with current training methods.