just saw a presentation on tensorflow 3.0 and i'm still trying to figure out why they didn't just use pytorch like the rest of us