brrrrrm

38 ms·

361.
A Beginner’s Guide to eBPF Programming – Liz Rice(youtube.com)
1 points by brrrrrm 4y ago|1 comments
362.
by brrrrrm 4y ago
> apply the cognitive biases model to algorithms which have superhuman performance in various games Could you give an example of this?
363.
by brrrrrm 4y ago
> comma is a 13 person company. We are ruthless about hiring the best and obsessed with minimizing complexity. How many is it now? Are you hiring?
364.
A visual guide to optimized matrix multiplication(twitter.com)
1 points by brrrrrm 4y ago|0 comments
365.
by brrrrrm 4y ago
How is ECMAScript 6 (high level language) related to WebAssembly (low level target)?
366.
by brrrrrm 4y ago
Edit: this is wrong. For posterity my original comment was: “From what I understand, tail calls can always (?) be lowered to while loops[1], which are expressible in WASM. 1. https://en.wikipedia.org/wiki/Tail_call#Rela
367.
by brrrrrm 4y ago
You have two sealed envelopes, one carrying a red card and the other a black. You shuffle the envelopes randomly and give one to your friend who goes to Mars. Once there, your friend opens their envelope and immediately knows the color of
368.
by brrrrrm 4y ago
I humbly submit another hypothetical that might help explain this idea: A drawing of a plane and the claim that “this thing can fly”. Unless you capture extremely subtle aspects about the materials and curvature of components, the depicted
369.
by brrrrrm 4y ago
“a system for electronic transactions without relying on trust.“ - From the original Bitcoin paper.
370.
Mlvt: Machine Learning Visualization Tools in the Terminal(github.com)
3 points by brrrrrm 4y ago|0 comments
371.
by brrrrrm 4y ago
For inference, models should be compiled down directly to WASM or WebGPU or whatever, right? The driving language really shouldn't matter at the end of the day. Unless you've got massive compute bound transformers or old-school fu
372.
by brrrrrm 4y ago
I think it predates that nomenclature. Google around for "perceptrons" instead. https://www.cs.utexas.edu/~lin/papers/hpca01.pdf
373.
by brrrrrm 4y ago
today, learned heuristics have a couple of pitfalls that make them hard to add to such systems 1. they are usually hard to run efficiently 2. they are usually hard to explain The former is definitely changing with low precision formats like
374.
by brrrrrm 4y ago
you might be interested in CompilerGym https://compilergym.com
375.
Bun is 60% faster than node on an M1 for a Fibonacci number implementation(twitter.com)
2 points by brrrrrm 4y ago|0 comments
376.
by brrrrrm 4y ago
Some of the phrasing is quite fascinating! E.g. “kidnap personal information”
377.
Proposal: Torch.js, a JavaScript front end for PyTorch(dev-discuss.pytorch.org)
2 points by brrrrrm 4y ago|0 comments
378.
by brrrrrm 4y ago
this is actually a brand new proposal! If you like it or have any comments, you can find more information here: https://dev-discuss.pytorch.org/t/proposal-torch-js-a-javasc...
379.
by brrrrrm 4y ago
Would you mind sharing which device you're using? I'm still working on making the first compilation faster!
380.
Mnist Demo Compiled On-the-Fly to WebAssembly in the Browser(mnist-demo.glitch.me)
52 points by brrrrrm 4y ago|21 comments
381.
by brrrrrm 4y ago
For what kind of work?
382.
by brrrrrm 4y ago
About 10x, right?
383.
SIMD Library for Evaluating Elementary Functions, Vectorized Libm and DFT(github.com)
2 points by brrrrrm 4y ago|0 comments
384.
by brrrrrm 4y ago
Not mentioned in the list of users is SLEEF ( https://github.com/shibatch/sleef ), which provides fast approximations for various elementary functions. (It generates coefficients for the approximations with mpfr) SLEEF
385.
by brrrrrm 4y ago
If the tensor cores are used, the 3090 can hit 300Tflops. I don’t think the M1 can even come close to that (21Tflops for the ultra)
386.
by brrrrrm 4y ago
pointwise isn’t a case of grouped conv, they’re orthogonal ideas. You can fuse grouped convs (depthwise is a special case of grouped convs) into preceding or following layers. Maybe JAX can do this already? No clue if any library offers su
387.
by brrrrrm 4y ago
Ah yes, I misunderstood your original comment
388.
by brrrrrm 4y ago
to clarify: I am comparing FP16 performance, which both the GPU and AMX have native support for. FP64 is also supported by AMX, making it quite an impressive region of silicon.
389.
by brrrrrm 4y ago
have you actually benchmarked that? I think (someone please correct me if I'm way off here) the AMX instructions can hit ~2.8tflops (fp16) per co-processor and there are 2 on the 7-core M1. That's 5.6tflops vs the 4.6tflops the
390.
by brrrrrm 4y ago
> We're not getting smell. why not? https://www.sciencedaily.com/releases/2018/11/181127092549.h...

More ›