Evaluating LLMs with fewer examples

Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.

Basic Sources for MIT 6.824 Distributed Systems Class

DeepSpec: a full-stack codebase for training and evaluating speculative decoding algorithms

A Tmux session manager, with preview, fuzzy finding, and MORE

An agentic skills framework & software development methodology that works.

Ralph is an autonomous AI agent loop that runs repeatedly until all PRD items are complete.

Manage your dotfiles across multiple diverse machines, securely.

Sequential Monte Carlo Speculative Decoding

Evaluate and Enhance Your LLM Deployments for Real-World Inference Needs

A unified library for building, evaluating, and storing speculative decoding algorithms for LLM inference in vLLM

A PyTorch native library for training speculative decoding models

AI agents running research on single-GPU nanochat training automatically

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

Machine Learning Systems

RLinf: Reinforcement Learning Infrastructure for Embodied and Agentic AI

Evaluate and improve models and agents using environments

Figure sizes, font sizes, fonts, and more configurations at minimal overhead. Fix your journal papers, conference proceedings, and other scientific publications.

My learning notes for ML SYS.

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflo…

Declarative visualization library for Python

PyTorch-native post-training at scale

A project to improve skills of large language models

A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downs…

The RL Bridge for LLM-based Agent Applications. Made Simple & Flexible.

SGLang is a high-performance serving framework for large language models and multimodal models.

Simple speculative decoding technique, integrated in vLLM and transformers

Train speculative decoding models effortlessly and port them smoothly to SGLang serving.

gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI

Renderer for the harmony response format to be used with gpt-oss