Sep 4, 2018 · 1 min read
Awesome article! Question, using the above architecture, and tricks, about how long will it take to train cifar10, on say frogs, or cars, given some sort of average-ish gpu, eg a 1080, or an aws g3 instance, or similar-ish?
What about for mnist?
What about for multiple cifar categories?
(just to have an approximate idea)
