To my satisfaction (reading quickly without doing exercises doesn’t count, but no need to do literally every problem if some seem redundant to me. Also i reserve the right to skip the JAX chapters if i think they are not useful to me)
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ116 | |
| 2 | Ṁ93 | |
| 3 | Ṁ26 | |
| 4 | Ṁ19 | |
| 5 | Ṁ4 |
I'm not sure how much knowing the math behind neutral networks contributes to practical knowledge about what they can do, or how to build one. The abstractions you get with tensor operations in pytorch are pretty good, you don't really have to know how they work to build any model with any training goal.
And that math looks really boring to study.
Karpathy has a pretty good set of videos on building and training transformer models. He explains how they work and train without getting into too much of the math underlying.
https://youtu.be/kCc8FmEb1nY?si=l4jZO-SKK33V5TdU