A toolkit for interpreting and analyzing neural networks (vision)

Quick calculation toolkit for x-ray imaging

Entropy Based Sampling and Parallel CoT Decoding

Scaling law by data manifold.

Transformer with Mu-Parameterization, implemented in Jax/Flax. Supports FSDP on TPU pods.

A Citation Manager and Zotero Integration for RemNote! Cite research all within your knowledge base!

Universal Notation for Tensor Operations in Python

An open-source implementation of Scaling Laws for Neural Language Models using nanoGPT

Score LLM pretraining data with classifiers

High accuracy RAG for answering questions from scientific documents with citations

Urai AE, Doiron B, Leifer AM & Churchland AK (2022) Large-scale neural recordings call for new insights to link brain and behavior. Nature Neuroscience

🧬 Generative modeling of regulatory DNA sequences with diffusion probabilistic models 💨

Trainable, memory-efficient, and GPU-friendly PyTorch reproduction of AlphaFold 2

Monte Carlo tree search in JAX

RITA is a family of autoregressive protein models, developed by LightOn in collaboration with the OATML group at Oxford and the Debora Marks Lab at Harvard.

A case study of efficient training of large language models using commodity hardware.

Simple and efficient RevNet-Library for PyTorch with XLA and DeepSpeed support and parameter offload

Nested Hierarchical Transformer https://arxiv.org/pdf/2105.12723.pdf

Official Pytorch Implementation of: "ImageNet-21K Pretraining for the Masses"(NeurIPS, 2021) paper

Your fruity companion for transformers

CLU lets you write beautiful training loops in JAX.

The largest collection of PyTorch image encoders / backbones. Including train, eval, inference, export scripts, and pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (V…

Contrastive Language-Audio Pretraining

JAX-based neural network library

Flax is a neural network library for JAX that is designed for flexibility.

Optax is a gradient processing and optimization library for JAX.

JAX + Haiku package for the discrete VAE used for DALL·E.