This isn't obvious. Most people remember to do data prefetching, but few do asynchronous logging. Keras `fit()` does both out of the box automatically, regardless of your data source. This gets you fully compact train_step stacking, like here — zero idle time.

→ View original post on X — @fchollet