1:01:08 Pretty sure this ODQA task that he mentions also allows you to reference documents -- in fact, you're supposed to. I think he's meant to be referring to "Closed-Book Question Answering"
19:30 I don't really understand what he's saying about how jointly training the embedding and model parameters leads to better initialization of parameters -- can someone clue me in?
Really appreciated how he repeated every student's question before answering. It makes it a lot easier for the remote learners to follow.
Agreed, and now I think Columbia university students will really enjoy his class as he becomes AP there.
Really nice how he articulates his thoughts in a short amount of time!
This lecture is great!
So well explained
awesome
Keep it up!
1:01:08
Pretty sure this ODQA task that he mentions also allows you to reference documents -- in fact, you're supposed to. I think he's meant to be referring to "Closed-Book Question Answering"
19:30
I don't really understand what he's saying about how jointly training the embedding and model parameters leads to better initialization of parameters -- can someone clue me in?
Wow you rarely (me: never) see a PhD student teach that well. The celebrated Karpathy doesn‘t come even close 😮
Agree whole-heartedly with this sentiment! Thank you to Stanford for making these lectures publicly available!
Karpathy is an engineer, not a teacher.
this guy is dancing in the classroom😄
51:17 Even the best universities in the world have projector related issues 🙃
💙💙💙
1:02:48
goldmine