JAX backend for SGL

This work documents a systematic investigation of the Julia programming language on Google Tensor Processing Unit (TPU) hardware, using Reactant.jl and the XLA compiler as the execution backend.

Open source Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents. Built for multitasking, organization, and programmability.

Official Codebase for "Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights" (ICML 2026 Spotlight)

小红书 CLI — 搜索、分析、自动化 Xiaohongshu content. Built for AI agents.

BitDance & UniWeTok: Open-source autoregressive model with binary visual tokens. A research project for building powerful multimodal autoregressive model.

Framework for evolving code and prompts inspired by Darwinian evolution

RL reasoning baselines

An open-source, code-first Python toolkit for building, evaluating, and deploying sophisticated AI agents with flexibility and control.

AllenAI's post-training codebase

A curated list of Decision Transformer resources (continually updated)

A JAX backend for Apple Metal Performance Shaders (MPS), enabling GPU-accelerated JAX computations on Apple Silicon.

Quantization, kernels, runtime and inference engine for mobiles, wearables, smart home and robots.

Voice tasks from Push iOS app → Claude Code & OpenAI Codex

Differentiable Hierarchical Visual Tokenization

StarVector is a foundation model for SVG generation that transforms vectorization into a code generation task. Using a vision-language modeling architecture, StarVector processes both visual and te…

High Scale Checkpointing Replicator

slime is an LLM post-training framework for RL Scaling.

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…

NanoGPT (124M) in 90 seconds

Accelerate, Optimize performance with streamlined training and serving options with JAX.

Minimal yet performant LLM examples in pure JAX

Raster to Vector Graphics Converter

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

Minimal, lightweight JAX implementations of popular models.

ret-sync is a set of plugins that helps to synchronize a debugging session (WinDbg/GDB/LLDB/OllyDbg2/x64dbg) with IDA/Ghidra/Binary Ninja disassemblers.

Simple, hackable transformer training & inference, in bare-metal JAX

Tokamax: A GPU and TPU kernel library.