Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more

Efficient Triton Kernels for LLM Training

Scenic: A Jax Library for Computer Vision Research and Beyond

Differentiable SDE solvers with GPU support and efficient sensitivity analysis.

Matplotlib style sheets to nicely format figures for scientific papers, thesis and presentations while keeping them fully editable in Adobe Illustrator.

Library for reading and processing ML training data.

Contrastive Learning of Structured World Models

Pretrained deep learning models for Jax/Flax: StyleGAN2, GPT2, VGG, ResNet, etc.

CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning

Code for the paper: Complex-Valued Autoencoders for Object Discovery

Code for the paper: Rotating Features for Object Discovery

🚀 A powerful library for efficient training of Neural Fields at scale.