ML Benchmarks

This repository contains various Machine Learning models implemented using different languages, backends and optimization techniques. The purpose is to benchmark the performance characteristics of alternative technologies.

Technologies

Currently Supported:

• WebGPU
• Rust
• Deno (TypeScript)
• Tensorflow.js (webgpu and cpu backend, but webgpu doesnt converge)

To be Added:

• TensorFlow
• WASM
• PyTorch
• Jax
• Haskell
• Lisp

Models and Benchmarks

• MNIST
• Multi Layer Perceptron

Useage

Run the benchmarks using:

deno run --unstable --allow-all src/index.ts

cargo run --release