Most important quote of the post: "but my personal experience is that the state of the art approach to exploring a nice and wide space of models and hyperparameters is to use an intern :). Just kidding."

"The first step to training a neural net is to not touch any neural net code" -- so very Zen.

:) "One time I accidentally left a model training during the winter break and when I got back in January it was SOTA (“state of the art”)."

❤️here's a yt link for the webinar youtube.com/watch?v=1jnKHP… and thank you for the recipe! I'll have to cook up a much longer presentation to cover this one.

Really love this quote: “Many times it is possible to feel the network “struggle” to fit your data if it wiggles too much in some way, revealing instabilities.”

1) I took cs231n and remember u standing at the auditorium entrance to greet us :) Thanks for that! 2) Can we get your permission to translate this to Arabic? I work with a team of volunteers to translate great articles and this would be very useful for non english speakers!

Hi Andrej, great presentation on Monday. Anything you can share when the software on V3 matches current software on V2.5? I know Elon said a few months yesterday. Excited to see what your team can do!

"init well." I've been trying to use learning rate finder techniques such as: towardsdatascience.com/estimating-opt…. I think the init well part was clearly a missing piece !

Thanks a lot for having the time to share this! I hope you don’t get discouraged by Tesla’s haters. You’re changing the world. You can disrupt and you will!