JAX in JavaScript – ML library for the web, running on WebGPU & Wasm

MoE training for Me and You and maybe other people

A natively parallel dataloader for Python, written in Rust. Serving data at GB/s speeds, while covering aspect ratio bucketing, crop and resize for image ML workloads.

Real-time terminal monitor for InfiniBand networks - htop for high-speed interconnects

Train Models Contrastively in Pytorch

encoder models with same training harness as nanogpt

Our library for RL environments + evals

CLIP-like model evaluation

jsgrad is a dependency-free ML library in Typescript for model inference and training with support to WebGPU and other runtimes.

Simplifying reinforcement learning for complex game environments

KernelBench: Can LLMs Write GPU Kernels? - Benchmark + Toolkit with Torch -> CUDA (+ more DSLs)

NanoGPT (124M) in 2 minutes

Bringing BERT into modernity via both architecture changes and scaling

code for training & evaluating Contextual Document Embedding models

Fast State-of-the-Art Static Embeddings

GPT4All: Run Local LLMs on Any Device. Open-source and available for commercial use.

Train VAE like a boss

A browser extension to help with WebGPU debugging

Virtual whiteboard for sketching hand-drawn like diagrams

Code for the MTEB Arena

Helpful tools and examples for working with flex-attention

Zoomable, animated scatterplots in the browser that scales over a billion points

Embedding Vector Oriented Clustering

run embeddings in MLX

A Python library for fast and easy access to genomic resources such as sequence, data tracks, and annotations

Minimal sharded dataset loaders, decoders, and utils for multi-modal document, image, and text datasets.