SimSIMD
gpu.js
SimSIMD | gpu.js | |
---|---|---|
15 | 9 | |
715 | 14,953 | |
- | 0.3% | |
9.6 | 0.0 | |
21 days ago | 2 months ago | |
C | JavaScript | |
Apache License 2.0 | MIT License |
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.
SimSIMD
- Deep Learning in JavaScript
-
From slow to SIMD: A Go optimization story
For other languages (including nodejs/bun/rust/python etc) you can have a look at SimSIMD which I have contributed to this year (made recompiled binaries for nodejs/bun part of the build process for x86_64 and arm64 on Mac and Linux, x86 and x86_64 on windows).
[0] https://github.com/ashvardanian/SimSIMD
-
Python, C, Assembly β Faster Cosine Similarity
Kahan floats are also commonly used in such cases, but I believe there is room for improvement without hitting those extremes. First of all, we should tune the epsilon here: https://github.com/ashvardanian/SimSIMD/blob/f8ff727dcddcd14...
As for the 64-bit version, its harder, as the higher-precision `rsqrt` approximations are only available with "AVX512ER". I'm not sure which CPUs support that, but its not available on Sapphire Rapids.
- Beating GCC 12 - 118x Speedup for Jensen Shannon Divergence via AVX-512FP16
- Show HN: Beating GCC 12 β 118x Speedup for Jensen Shannon D. Via AVX-512FP16
- SimSIMD v2: Vector Similarity Functions 3x-200x Faster than SciPy and NumPy
-
Show HN: SimSIMD vs. SciPy: How AVX-512 and SVE make SIMD cleaner and ML faster
I encourage one to merge into e.g. {NumPy, SciPy, }; are there PRs?
Though SymPy.physics only yet supports X,Y,Z vectors and doesn't mention e.g. "jaccard"?, FWIW: https://docs.sympy.org/latest/modules/physics/vector/vectors... https://docs.sympy.org/latest/modules/physics/vector/fields.... #cfd
include/simsimd/simsimd.h: https://github.com/ashvardanian/SimSIMD/blob/main/include/si...
conda-forge maintainer docs > Switching BLAS implementation:
-
SimSIMD v2: 3-200x Faster Vector Similarity Functions than SciPy and NumPy
Hello, everybody! I was working on the next major release of USearch, and in the process, I decided to generalize its underlying library - SimSIMD. It does one very simple job but does it well - computing distances and similarities between high-dimensional embeddings standard in modern AI workloads.
- Comparing Vectors 3-200x Faster than SciPy and NumPy
-
Show HN: U)Search Images demo in 200 lines of Python
Hey everyone! I am excited to share updates on four of my & my teams' open-source projects that take large-scale search systems to the next level: USearch, UForm, UCall, and StringZilla. These projects are designed to work seamlessly together, end-to-endβcovering everything from indexing and AI to storage and networking. And yeah, they're optimized for x86 AVX2/512 and Arm NEON/SVE hardware.
USearch [1]: Think of it as Meta FAISS on steroids. It's now quicker, supports clustering of any granularity, and offers multi-index lookups. Plus, it's got more native bindings than probably all other vector search engines combined: C++, C, Python, Java, JavaScript, Rust, Obj-C, Swift, C#, GoLang, and even slightly outdated bindings for Wolfram. Need to refresh that last one!
UForm v2 [2]: Imagine a much smaller OpenAI CLIP but more efficient and trained on balanced multilingual datasets, with equal exposure to languages from English, Chinese, and Hindi to Arabic, Hebrew, and Armenian. UForm now supports 21 languages, is so tiny that you can run it in the browser, and outputs small 256-dimensional embeddings. Perfect for rapid image and video searches. It's already available on Hugging-Face as "unum-cloud/uform-vl-multilingual-v2".
UCall [3]: It started as a FastAPI alternative focusing on JSON-RPC (instead of REST protocols), offering 70x the bandwidth and 1/50th the latency. It was good but not enough, so we've added REST and TLS support, broadening its appeal. I've merged that code, and it is yet to be tested. Early benchmarks suggest that we still hit the same 150'000-250'000 requests/s on a single CPU core in Python by reusing HTTPS connections.
StringZilla [4]: This project lets you sift through multi-gigabyte or terabyte strings with minimal use of RAM and maximal use of SIMD and SWAR techniques.
All these projects are engineered for scalability and efficiency, even on tight budgets. Our demo, for instance, works on hundreds of gigabytes of images using just a few gigabytes of RAM and no GPUs for AI inference. That is a toy example with a small, noisy dataset, and I look forward to showing a much larger setup. Interestingly, even this tiny setup illustrates issues common to UForm and much larger OpenAI CLIP models - the quality of Multi-Modal alignment [5]. It also shows how different/accurate the search results are across different languages. Synthetic benchmarks suggest massive improvements for some low-resource languages (like Armenian and Hebrew) and more popular ones (like Hindi and Arabic) [6]. Still, when we look at visual demos like this, I can see a long road ahead for us and the broader industry, making LLMs Multi-Modal in 2024 :)
All of the projects and the demo code are available under an Apache license, so feel free to use them in your commercial projects :)
PS: The demo looks much nicer with just Unsplash dataset of 25'000 images, but it's less representative of modern AI datasets, too small, and may not be the best way to honestly show our current weaknesses. The second dataset - Conceptual Captions - is much noisier, and quite ugly.
[1]: https://github.com/unum-cloud/usearch
gpu.js
-
Deep Learning in JavaScript
You might already be familiar, but a GPU.js backend can provide some speedups via good old WebGL -- no need for WebGPU just yet!
[0]: https://github.com/gpujs/gpu.js/
-
Show HN: Shadeup β A language that makes WebGPU easier
Very cool project.
I learned WebGL three years ago but before I dove into the underlying concepts I used GPU.js [1] to quickly prototype my project. Eventually, the abstraction prevented necessary performance optimizations so I switched to vanilla GLSL and these vanilla GLSL "shaders" were initially ejected from GPU.js.
Writing JS code then looking at the generated WebGPU output is a great way to get familiar with WebGPU. Thanks for this.
[1] https://github.com/gpujs/gpu.js/
-
Gpu.js: GPU Accelerated JavaScript
I used this library on my project but I think it's no longer maintained. I PRed a fix for buggy atan2 over a year ago and no movement [1]. I do highly recommend it if you're a web developer interested in harnessing parallel processing.
[1] https://github.com/gpujs/gpu.js/pull/683
-
Brain.js: GPU Accelerated Neural Networks in JavaScript
Thanks for pointing this out. I've submitted a PR to resolve this: https://github.com/gpujs/gpu.js/issues/757
That being said, if you're not building from source (you're running an LTS version of node on a supported platform), you don't need to worry about python or many of the build deps.
- GPU.js
- For what projects, Nodejs is an absolute No No?
What are some alternatives?
kuzu - Embeddable property graph database management system built for query speed and scalability. Implements Cypher.
numjs - Like NumPy, in JavaScript
nsimd - Agenium Scale vectorization library for CPUs and GPUs
headless-gl - π Windowless WebGL for node.js
numpy-feedstock - A conda-smithy repository for numpy.
math-clamp - Clamp a number
mkl_random-feedstock - A conda-smithy repository for mkl_random.
aladino - π§ββοΈ Your magic WebGL carpet
usearch - Fast Open-Source Search & Clustering engine Γ for Vectors & π Strings Γ in C++, C, Python, JavaScript, Rust, Java, Objective-C, Swift, C#, GoLang, and Wolfram π
math-sum - Sum numbers
xtensor-fftw - FFTW bindings for the xtensor C++14 multi-dimensional array library
Brain.js - π€ GPU accelerated Neural networks in JavaScript for Browsers and Node.js