https://github.com/Game4all/brainz

🧠 Braaaaaaaaainnnnnss 🧟‍♂️ - half finished zig ml lib
https://github.com/Game4all/brainz

Last synced: about 1 month ago
JSON representation

🧠 Braaaaaaaaainnnnnss 🧟‍♂️ - half finished zig ml lib

• Host: GitHub
• URL: https://github.com/Game4all/brainz
• Owner: Game4all
• Created: 2024-04-27T15:38:04.000Z (8 months ago)
• Default Branch: master
• Last Pushed: 2024-10-01T20:58:52.000Z (3 months ago)
• Last Synced: 2024-11-21T08:42:02.402Z (about 1 month ago)
• Language: Zig
• Homepage:
• Size: 232 KB
• Stars: 4
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        


    
brainz

    Braaaaaaaaaaaaaiiiiiiiinnnnnnnnsssss 🧠🧟‍♂️

    


    A small tensor library with just enough operators for Deep Learning written in pure Zig

    




> **Warning**

> This library is a work in progress. Design considerations aren't set in stone.

This library implements a `Tensor` primitive and a subset of operators to design, train and run deep learning models using Zig, on any platform that runs Zig code.

See [ops.zig](src/ops.zig) for a list of currently implemented operators. Also note that making your own is quite straightforward too!

The library provides an abstraction for multithreading / accelerator support with _explicit synchronization_ through the `Device` interface. Operators take a device as parameter and are dispatched in a non-blocking way. 

**Synchronization is totally explicit and it is up to the programmer to between insert sync points between dependent computations**.

A multi-threaded CPU device is currently available, but we could maybe get some juicy GPU accel someday to go zoomin (soon-ish™).

## Installation

After adding the project to your `build.zig.zon`, add to your `build.zig`

```

pub fn build(b: *std.Build) void {

    const target = b.standardTargetOptions(.{});

    const optimize = b.standardOptimizeOptions(.{});

    const brainz = b.dependency("brainz", .{

        .optimize = optimize,

        .target = target,

    });

    your_project_executable.root_module.addImport("brainz", brainz.module("brainz"));

}

```

## Getting started

 Please check the `examples/` directory to get started.

 Want something a little bigger to look at ? See [MNIST from scratch](https://github.com/Game4all/mnist-from-scratch)