JAX numpy killer

References:

Outline
Examples overview: (0:00​)
Source code review: (27:36)
Documentation reivew: (43:04)

[samersabri]

Thanks this covers a lot of things I didn't fully understand and explains it without any fluff. Not sure if you eventually figured it out, but the code you weren't sure about at min 5 is actually a mlp and the code assigns to the input because the output of the current layer in the current iteration of the loop becomes the input for the next layer/iteration. Notice how bad I am at simplifying and making things non-fluffy lol

[ChaiTimeDataScience]

55:48

Does anyone know the frameworks mentioned here? I couldn't catch the names and want to check them out.

This was another awesome video, Mark's videos are my new favorite weekend binges!

[orjihvy]

I guess, the idea is that even though tensorflow can do this as well, you don't have all the overhead that comes with tensorflow, if you need tools that can do automatic differentiation and some matrix manipulation that simply works well or better than Numpy.

[orjihvy]

may be the fonts on the screen are a bit too small

[pyatsyshin]

Mate, what the hell is happening at 12:33 with your result? It should be [3, 8, 18]. What's going on there and why do you call it "correct now" ??