CUDA.jl
18337
CUDA.jl | 18337 | |
---|---|---|
15 | 14 | |
1,133 | 189 | |
1.1% | 3.2% | |
9.5 | 5.7 | |
7 days ago | about 1 year ago | |
Julia | Jupyter Notebook | |
GNU General Public License v3.0 or later | - |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
CUDA.jl
-
Ask HN: Best way to learn GPU programming?
It would also mean learning Julia, but you can write GPU kernels in Julia and then compile for NVidia CUDA, AMD ROCm or IBM oneAPI.
https://juliagpu.org/
I've written CUDA kernels and I knew nothing about it going in.
- What's your main programming language?
-
How is Julia Performance with GPUs (for LLMs)?
See https://juliagpu.org/
-
Yann Lecun: ML would have advanced if other lang had been adopted versus Python
If you look at Julia open source projects you'll see that the projects tend to have a lot more contributors than the Python counterparts, even over smaller time periods. A package for defining statistical distributions has had 202 contributors (https://github.com/JuliaStats/Distributions.jl), etc. Julia Base even has had over 1,300 contributors (https://github.com/JuliaLang/julia) which is quite a lot for a core language, and that's mostly because the majority of the core is in Julia itself.
This is one of the things that was noted quite a bit at this SIAM CSE conference, that Julia development tends to have a lot more code reuse than other ecosystems like Python. For example, the various machine learning libraries like Flux.jl and Lux.jl share a lot of layer intrinsics in NNlib.jl (https://github.com/FluxML/NNlib.jl), the same GPU libraries (https://github.com/JuliaGPU/CUDA.jl), the same automatic differentiation library (https://github.com/FluxML/Zygote.jl), and of course the same JIT compiler (Julia itself). These two libraries are far enough apart that people say "Flux is to PyTorch as Lux is to JAX/flax", but while in the Python world those share almost 0 code or implementation, in the Julia world they share >90% of the core internals but have different higher levels APIs.
If one hasn't participated in this space it's a bit hard to fathom how much code reuse goes on and how that is influenced by the design of multiple dispatch. This is one of the reasons there is so much cohesion in the community since it doesn't matter if one person is an ecologist and the other is a financial engineer, you may both be contributing to the same library like Distances.jl just adding a distance function which is then used in thousands of places. With the Python ecosystem you tend to have a lot more "megapackages", PyTorch, SciPy, etc. where the barrier to entry is generally a lot higher (and sometimes requires handling the build systems, fun times). But in the Julia ecosystem you have a lot of core development happening in "small" but central libraries, like Distances.jl or Distributions.jl, which are simple enough for an undergrad to get productive in a week but is then used everywhere (Distributions.jl for example is used in every statistics package, and definitions of prior distributions for Turing.jl's probabilistic programming language, etc.).
-
C++ is making me depressed / CUDA question
If you just want to do some numerical code that requires linear algebra and GPU, your best bet would be Julia or Python+JAX.
-
Parallélisation distribuée presque triviale d’applications GPU et CPU basées sur des Stencils avec…
GitHub - JuliaGPU/CUDA.jl: CUDA programming in Julia.
- Why Fortran is easy to learn
-
Generic GPU Kernels
Should have (2017) in the title.
Indeed cool to program julia directly on the GPU and Julia on GPU and this has further evolved since then, see https://juliagpu.org/
-
Announcing The Rust CUDA Project; An ecosystem of crates and tools for writing and executing extremely fast GPU code fully in Rust
I'm excited to eventually see something like JuliaGPU with support for multiple backends.
-
[Media] 100% Rust path tracer running on CPU, GPU (CUDA), and OptiX (for denoising) using one of my upcoming projects. There is no C/C++ code at all, the program shares a single rust crate for the core raytracer and uses rust for the viewer and renderer.
That's really cool! Have you looked at CUDA.jl for the Julia language? Maybe you could take some ideas from there. I am pretty sure it does the same thing you do here, and they support any arbitrary code with the limitations that you cannot allocate memory, I/O is disallowed, and badly-typed code(dynamic) will not compile.
18337
- Hello I wanted to know what would be the best way to get started in Julia and artificial intelligence. I looked around alot of different languages and saw Julia was good for data science and for artificial intelligence but would like to know what would be good ways to just do it. Thank you
-
SciML/SciMLBook: Parallel Computing and Scientific Machine Learning (SciML): Methods and Applications (MIT 18.337J/6.338J)
This was previously the https://github.com/mitmath/18337 course website, but now in a new iteration of the course it is being reset. To avoid issues like this in the future, we have moved the "book" out to its own repository, https://github.com/SciML/SciMLBook, where it can continue to grow and be hosted separately from the structure of a course. This means it can be something other courses can depend on as well. I am looking for web developers who can help build a nicer webpage for this book, and also for the SciMLBenchmarks.
-
Why Fortran is easy to learn
I would say Fortran is a pretty great language for teaching beginners in numerical analysis courses. The only issue I have with it is that, similar to using C+MPI (which is what I first learned with, well after a bit of Java), the students don't tend to learn how to go "higher level". You teach them how to write a three loop matrix-matrix multiplication, but the next thing you should teach is how to use higher level BLAS tools and why that will outperform the 3-loop form. But Fortran then becomes very cumbersome (`dgemm` etc.) so students continue to write simple loops and simple algorithms where they shouldn't. A first numerical analysis course should teach simple algorithms AND why the simple algorithms are not good, but a lot of instructors and tools fail to emphasize the second part of that statement.
On the other hand, the performance + high level nature of Julia makes it a rather excellent tool for this. In MIT graduate course 18.337 Parallel Computing and Scientific Machine Learning (https://github.com/mitmath/18337) we do precisely that, starting with direct optimization of loops, then moving to linear algebra, ODE solving, and implementing automatic differentiation. I don't think anyone would want to give a homework assignment to implement AD in Fortran, but in Julia you can do that as something shortly after looking at loop performance and SIMD, and that's really something special. Steven Johnson's 18.335 graduate course in Numerical Analysis (https://github.com/mitmath/18335) showcases some similar niceties. I really like this demonstration where it starts from scratch with the 3 loops and shows how SIMD and cache-oblivious algorithms build towards BLAS performance, and why most users should ultimately not be writing such loops (https://nbviewer.org/github/mitmath/18335/blob/master/notes/...) and should instead use the built-in `mul!` in most scenarios. There's very few languages where such "start to finish" demonstrations can really be showcased in a nice clear fashion.
-
What are some interesting papers to read?
And why not take a course while you're at it.
- Composability in Julia: Implementing Deep Equilibrium Models via Neural Odes
- [2109.12449] AbstractDifferentiation.jl: Backend-Agnostic Differentiable Programming in Julia
-
Is that true?
Here's a good one. It's in Julia but it should do the trick. The main instructor is the most prolific Julia dev in the world.
-
[D] Has anyone worked with Physics Informed Neural Networks (PINNs)?
NeuralPDE.jl fully automates the approach (and extensions of it, which are required to make it solve practical problems) from symbolic descriptions of PDEs, so that might be a good starting point to both learn the practical applications and get something running in a few minutes. As part of MIT 18.337 Parallel Computing and Scientific Machine Learning I gave an early lecture on physics-informed neural networks (with a two part video) describing the approach, how it works and what its challenges are. You might find those resources enlightening.
-
[P] Machine Learning in Physics?
It's a very thriving field. If you are interested in methods research and want to learn some of the techniques behind it, I would recommend taking a dive into my lecture notes as I taught a graduate course at MIT, 18.337 Parallel Computing and Scientific Machine Learning, specifically designed to get new students onboarded into this research program.
- MIT 18.337J: Parallel Computing and Scientific Machine Learning
What are some alternatives?
LoopVectorization.jl - Macro(s) for vectorizing loops.
DataDrivenDiffEq.jl - Data driven modeling and automated discovery of dynamical systems for the SciML Scientific Machine Learning organization
cunumeric - An Aspiring Drop-In Replacement for NumPy at Scale
Vulpix - Fast, unopinionated, minimalist web framework for .NET core inspired by express.js
awesome-quant - A curated list of insanely awesome libraries, packages and resources for Quants (Quantitative Finance)
NeuralPDE.jl - Physics-Informed Neural Networks (PINN) Solvers of (Partial) Differential Equations for Scientific Machine Learning (SciML) accelerated simulation
cudf - cuDF - GPU DataFrame Library
SciMLTutorials.jl - Tutorials for doing scientific machine learning (SciML) and high-performance differential equation solving with open source software.
Tullio.jl - ⅀
GPUCompiler.jl - Reusable compiler infrastructure for Julia GPU backends.
BenchmarkTools.jl - A benchmarking framework for the Julia language