Stars
Skills for writing tilelang and debugging with CUDA toolkits.
works well with rocks; native Apple Silicon inference engine for capable and tiny models (ternary, Qwen) via MLX Swift. Hadamard-rotated KV cache, vision, tool calling, OpenAI + Anthropic APIs
Fast, standardized, and easy-to-use interpretability engine.
PyTorch building blocks for the OLMo ecosystem
Compile programs directly into transformer weights. Includes a 2D convex-hull KV cache with O(log n) inference.
Use as many MCP servers as you want while minimizing context usage. A code mode MCP server gateway driven with Lua 🌙
Python bindings to the Rust rpds crate for persistent data structures
RikkaHub is an Android APP that supports for multiple LLM providers.
Claude Discord bot rewritten in Elixir -- now public :3
High-performance multi-path covert channel over DNS
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…
AHN: Artificial Hippocampus Networks for Efficient Long-Context Modeling
Pure C++ implementation of several models for real-time chatting on your computer (CPU & GPU)
FlexAttention based, minimal vllm-style inference engine for fast Gemma 2 inference.
A minimal tensor processing unit (TPU), inspired by Google's TPU V2 and V1
rl from zero pretrain, can it be done? yes.
An open-source AI coding agent that lives in your terminal.
Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.
⚡️SwanLab - an open-source, modern-design AI training tracking and visualization tool. Supports Cloud / Self-hosted use. Integrated with PyTorch / Transformers / verl / LLaMA Factory / ms-swift / U…
🔥 A minimal training framework for scaling FLA models
Official repository for "BLEUBERI: BLEU is a surprisingly effective reward for instruction following"
A fast type checker and language server for Python
Code for collecting, processing, and preparing datasets for the Common Pile