PyTorch implementation of JiT https://arxiv.org/abs/2511.13720

Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization.

Open-Source Models for VLA Arena

Distributed MoE in a Single Kernel [NeurIPS '25]

Minimal PyTorch implementation of TP, SP, FSDP and sharded-EMA

CUTLASS and CuTe Examples

🔥 LLM-powered GPU kernel synthesis: Train models to convert PyTorch ops into optimized Triton kernels via SFT+RL. Multi-turn compilation feedback, cross-platform NVIDIA/AMD, Kernelbook + KernelBench

Minimum implementation of EDM (Elucidating the Design Space of Diffusion-Based Generative Models) on cifar10 and mnist

Code for "What really matters in matrix-whitening optimizers?"

Fast CUDA matrix multiplication from scratch

minimal Energy-based transformer

A foundation model to learn multiple physical systems at once

Code for "Transitive RL: Value Learning via Divide and Conquer"

🎨 Native AI image generation for Apple Silicon with Qwen-Image. Lightning LoRA acceleration for fast 4–8 step runs. Zero Docker, just works.

Chimera: State Space Models Beyond Sequences

Library for FO Optimization

torchax is a PyTorch frontend for JAX. It gives JAX the ability to author JAX programs using familiar PyTorch syntax. It also provides JAX-Pytorch interoperability, meaning, one can mix JAX & Pytor…

TPU inference for vLLM, with unified JAX and PyTorch support.

The official implementation of "Dual Goal Representations"

The best ChatGPT that $100 can buy.

StreamingVLM: Real-Time Understanding for Infinite Video Streams

The simplest, fastest repository for training/finetuning small-sized VLMs.

Pytorch Implementation (unofficial) of the paper "Mean Flows for One-step Generative Modeling" by Geng et al.

A JAX-native LLM Post-Training Library

Fast Diffusion Models with Transformers

Large multi-modal models (L3M) pre-training.