ffriend

11 ms·

1.
by ffriend 2y ago
There seem to be two distinct topics: 1) async programming vs. threading 2) infectious async/await syntax Async programming is great. Coroutines are a powerful tool, both for expressing your ideas more clearly and for improving perform
2.
by ffriend 2y ago
Sure, knowing the basics of LLM math is necessary. But it's also _enough_ to know this math to fully grasp the code. There are only 4 concepts - attention, feed-forward net, RMS-normalization and rotary embeddings - organized into a cl
3.
by ffriend 2y ago
It's also worth mentioning that the original implementation by Meta is only 300 lines of very readable code [1]. [1]: https://github.com/meta-llama/llama3/blob/main/llama/model.p...
4.
by ffriend 2y ago
JAX requires a bit more work to maintain fixed-size buffers as required by XLA, especially in case of caching and rotary embeddings. But yeah, overall the code can be pretty similar [1]. [1]: https://github.com/dfdx/fab
5.
by ffriend 8y ago
> there's nothing in the language that prevents this from working with the autograd package, except no one's taken the time to implement it ( https://github.com/HIPS/autograd/issues/47 ). I believe
6.
by ffriend 8y ago
I have quite limited experience with Cython and tried Numba just a couple of times, but I'm curious how much would it take to rewrite one of my Julia libraries to them. The library is for reverse-mode automatic differentiation, but let
7.
by ffriend 8y ago
Almost certainly yes. Although it doesn't mean it's the best choice for your specific task. Python had a long history of web server development. I think, aiohttp may be considered state of the art now. Let's measure its perfo
8.
by ffriend 9y ago
Some time ago I implemented a library [1] similar to this tool. The tricky part is that derivatives quickly exceed 2 dimensions, e.g. derivative of a vector output w.r.t. input matrix is a 3D tensor (e.g. if `y = f(X)`, you need to find der
9.
by ffriend 9y ago
If I want to export my own computational graph to ONNX, what is the first place I should look at? Do you know about any documentation or reference implementation of the format?
10.
by ffriend 9y ago
> I don’t understand their vendetta against the graph, which is a powerful abstraction that lets you choose different backend [...] You don't really need a graph to support different backends. One popular approach is to have differe
11.
by ffriend 9y ago
I could have not noticed use of swap in the previous test, so I repeated it on a Linux box and 1e8 numbers (instead of 1e9). Julia took 100.583ms while Python 207ms (probably due to double reading of the array). So I guess adding 1e9 number
12.
by ffriend 9y ago
BLAS really shines when you do matrix multiplication, for element-wise operations the best you can do is to add numbers using SIMD instructions or put the load to GPU, and most numeric libraries already when possible. The benchmark about se
13.
by ffriend 9y ago
Even better would be to write (see dot vectorization [1]): C .= A .+ B Benchmarks for 3 matrices of size 1000x1000: julia> using BenchmarkTools julia> @benchmark C = A + B BenchmarkTools.Trial: memory es
14.
by ffriend 9y ago
> The author understands your perspective but he's deliberately using a different one. In this case the way the author shows it isn't the best one: he modifies Python code to be more realistic - that's ok, but doesn't
15.
by ffriend 9y ago
The two language problem is more specific to scientific computing where literally every popular library is written in more then one language - fast C/C++/Fortran and convenient high-level R/Matlab/Python. NumPy, SciPy, C
16.
by ffriend 9y ago
As far as I know, storing personal data - including photo, name, email and sometimes even IP address - without explicit and clear consent is strictly forbidden in most countries, at least in EU.
17.
by ffriend 9y ago
If we are talking about professional actors trying to trick the tracker, then yes, it should be pretty hard to design software to overcome it. But most people aren't that good, and although they can mislead their friends or collegues,
18.
by ffriend 9y ago
Well, I definitely do have other applications for it. For example, I know that similar software has been used in labs to estimate people's reaction to videos and game features, in mobile applications to improve interaction with a user,
19.
by ffriend 9y ago
> I would feel bad about it and I think the person should ask my permission first. But if that person just memorizes customer reactions to understand how people on average react to particular products or actions, that's ok, right? B
20.
by ffriend 9y ago
I'm curious what's the boundary between ethical and unethical. People constantly analyze each other's mood and it's perceived positively. But doing the same thing massively using automated tools is often considered inapp
21.
by ffriend 9y ago
But isn't A/B testing doing the same thing? If it's different, what's the key difference between analyzing facial expression in a shop and analyzing user interaction on a site (given that both have a warning about data c
22.
by ffriend 9y ago
I see 2 separate issues here: 1. Accurate collection of facial features. Illumination, occlusions, head rotation, etc. may seriously affect accuracy, but this is exactly our main focus right now. We are at the very start of the process, yet
23.
by ffriend 9y ago
> I don't like having my emotions manipulated What about collecting statistics to make better decisions? Let's say, you go to your favorite jeans store, but find out that current collection is disgusting. Does it sound ok for y
24.
by ffriend 9y ago
Thanks for the detailed comment. Couple clarifying questions if you don't mind: - do you know that loyalty cards are often used in stores to collect customer data (a kind of offline cookie)? do you consider it a bad/dangerous&#x2F
25.
by ffriend 9y ago
As someone working on a similar project (specifically, emotion recognition) I'm highly interested to hear how such a product should look like to be not considered unethical. So far from the comments I see that: - it should be made clea
26.
by ffriend 10y ago
Fortran, Mathematica, Matlab, R, Lua - all start indexing from 1 (you can find an incomplete list in [1]). For users coming from these languages (which is a large part of the Julia community) the cognitive load would be in the opposite dire
27.
by ffriend 10y ago
> Is it more elegant I'd say it's more readable. Here's why I like to read Julia code more than Python code (and Python code is normally already pretty readable): * multiple dispatch in Julia allows to write shorter functi
28.
by ffriend 10y ago
> But my ruling for libraries is anyway: use as few as possible. And this is the point - if you want to be productive, you have to use _may_ libraries. Unfortunately, Scala's way is not to carry about compatibility, making it extrem
29.
by ffriend 11y ago
You can turn garbage collection on and off using `gc_enable` function.
30.
by ffriend 11y ago
True in Haskell, but not e.g. in Scala. In Scala, pattern matching is done using partial functions that are defined only for certain patterns. E.g. it's totally legal to write: obj match { case Foo(x) => ... case Bar(y,

More ›