On behalf of all humans: Thank you Andrej!     See Less

i did it

[pound pound pound]
derives from a mandate from the masses, accompanied by force!
TIM: I warned you! Oh, no, you! But did you knew it all,
LAUNCELOT: Ah. Let     See More ssing me in my liege.
W I warned you listen to beat us. Come on then.
[chord]
LEFT HEAD: Oh, Concorde!
DINGO: In the name of Assyria?
BEDEMIR: Yes, yes, remove the supports!
DENNIS: Witch!
HERBERT: Man!
CROWD: Burn her!    See Less

Come on Andrej!!! I can hardly wait for the next video! We love you man     See Less

This is a fantastically clear explanation, thank you so much!     See Less

You are impressively efficient. This was a blast to watch     See Less

hey andrej, this is where i'm lost. what's the point of linearly scaling down from channel embeddings of 32 to 16 head size? why don't we keep this at 32?     See Less

This man really has a gift in teaching stuff. Wish I was good at something but I suck at everything, (mediocrity is killing me). But hey, I completed the video feeling like I understood some     See More t's nice. Thank you Andrej    See Less

fyi for anyone on macbooks, change device = 'cuda' if if torch.cuda.is_available() else 'cpu' to

device = 'cuda' if torch.cuda.is_available() else 'mps     See More h.backends.mps.is_available() else 'cpu'

anyone else can make this change too, just makes it so the device checks for m1 gpu availability too before it reverts to cpu    See Less

Thank you - this bring sanity to the space by demonstrating exactly how it works. It's not 'magic'. It's not 'human'. It's code. thank you so much.     See Less

I only just noticed that this is set up in a perfect 2 column layout so a person can have the script/notebook they are working on side by side with yours and not have to jump around at all.     See More ;s clean and clutter free. That is some classy action, my deepest respect and gratitude.    See Less