pizza

46 ms·

811.
pytorch/torch/utils/_pytree.py(github.com)
1 points by pizza 4y ago|2 comments
812.
by pizza 4y ago
Does this work with jax jit?
813.
by pizza 4y ago
The Python ecosystem imo tries to get you from 0 to 1 a bit faster than the more performant alternatives with large userbases I’m really thinking of the C/C++ ecosystem here - os portability, compiler differences, cmake, xeus-cling vs
814.
by pizza 4y ago
einops, einsum, pytrees, named tensors, and xla/jit/tensor compilation (jax/tvm/taco) will hopefully improve things
815.
Ask HN: Computer-assisted prosthetic/workflow for math?
2 points by pizza 4y ago|0 comments
816.
How to temporarily remove an unwanted path name from %path% variable on Windows(stackoverflow.com)
1 points by pizza 4y ago|0 comments
817.
JPEG-XL Sample HDR Photos(people.csail.mit.edu)
8 points by pizza 4y ago|5 comments
818.
by pizza 4y ago
Would love to get proper hdr support on Ubuntu. Furthermore there doesn't seem to be display-stream compression support, so I can't get 5160x1440 @ 240 Hz :/ - "just" 3840x1080, which to be fair, is still many pixel
819.
NASA’s Webb takes star-filled portrait of Pillars of Creation(nasa.gov)
727 points by pizza 4y ago|191 comments
820.
by pizza 4y ago
I mean it kinda makes sense. When you're coding, you're constructing something. When you're debugging, you're deconstructing something. I feel like it's natural for people to take a sec to codeswitch, bc they were l
821.
by pizza 4y ago
4 years and counting, blimey
822.
tjax: A wide variety of tools for jax(neilgirdhar.github.io)
2 points by pizza 4y ago|0 comments
823.
Going Dutch(en.wikipedia.org)
3 points by pizza 4y ago|0 comments
824.
by pizza 4y ago
Here is a mini-dsl that I think has a really neat syntax: einops e.g. how to upsample an image 2x from einops import rearrange, repeat rearrange(repeat(img, 'h w c -> h w r1 r2 c', r1=2, r2=2), 'h w r1 r2 c -> (h
825.
by pizza 4y ago
Do you have a "More tweets" section under the main tweet? I have one, and I imagine that that is both generated on-the-fly and tailored to my browsing behavior. So it probably interrogates a larger system, adding latency. I'd
826.
by pizza 4y ago
It doesn't indicate either of those things. It's a great tool when you're working on complex code that just reaches the limit of your working memory/attention, and has often suggested good and clean improvements for me.
827.
Low Latency Audio API for Windows Drivers(learn.microsoft.com)
4 points by pizza 4y ago|0 comments
828.
Stable Diffusion Inference on iOS(github.com)
7 points by pizza 4y ago|0 comments
829.
by pizza 4y ago
Ah interesting. I was thinking along the lines of if you wanted to, say, package up an algorithm as a class, make a ton of these classes/algorithms, and then affix them in a bunch of ways. E.g. "find the algorithms that take in da
830.
by pizza 4y ago
This is awesome. Is there a way to get it to work with e.g. dataclasses-sql? I am thinking of something such as stuff: list[mydataclass] sql_val^ "select * from stuff where stuff.field = 2" over any kind of 'fmappab
831.
by pizza 4y ago
Boredom in the right hands leads to works of tremendous beauty
832.
by pizza 4y ago
I agree that pedagogy should be flexible. Not be static/constant, but, rather, a relationship that reacts to inputs from students; changing dynamically to suit the student's needs. It'd be a great thing if people could build
833.
by pizza 4y ago
> reminds me of what we could have achieved if this ingenuity had been applied in another domain It will; I think the reason we're seeing diffusion models applied to image generation first, is that it's a task that meshes well
834.
by pizza 4y ago
Why - you think it wasn't clever enough to be called clever hacking? :)
835.
by pizza 4y ago
> I remember reading that von Neumann was a fan of turning on the TV or radio to very loud volumes when he was working at home. Likely to achieve the same effect. There is a whole genre of YouTube videos that is just somebody walking in
836.
Functional Differential Geometry [pdf](web.mit.edu)
3 points by pizza 4y ago|0 comments
837.
by pizza 4y ago
Python programming is simple because I the programmer have to do less stuff. But also, to a really really crude approximation, some asm that does less stuff is both simpler, and faster, than some asm that does more stuff (depends on what it
838.
Geometric data analysis, beyond convolutions (2020) [pdf](jeanfeydy.com)
4 points by pizza 4y ago|0 comments
839.
by pizza 4y ago
The article is about having the correct environment set up - attacking the problem from the environment level. When I think of "hacking code you didn't write" one of the biggest hurdles for me is "inserting my changes to
840.
by pizza 4y ago
Invertibility, differentiability, adjointness, compositionality, dependency injection, instrumentation, and graph algorithms..

More ›