Stars
cuTile is a programming model for writing parallel kernels for NVIDIA GPUs
Tile-Based Runtime for Ultra-Low-Latency LLM Inference
CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning
Official Implementation of SAM-Decoding: Speculative Decoding via Suffix Automaton
A fast serialization and validation library, with builtin support for JSON, MessagePack, YAML, and TOML
Fast, correct Python JSON library supporting dataclasses, datetimes, and numpy
SWE-bench: Can Language Models Resolve Real-world Github Issues?
Spec-Bench: A Comprehensive Benchmark and Unified Evaluation Platform for Speculative Decoding (ACL 2024 Findings)
ArcticInference: vLLM plugin for high-throughput, low-latency inference
A debugging and profiling tool that can trace and visualize python code execution
Getting Started with Triton: A Tutorial for Python Beginners
🎒 Token-Oriented Object Notation (TOON) – Compact, human-readable, schema-aware JSON for LLM prompts. Spec, benchmarks, TypeScript SDK.
Qwen3Guard is a multilingual guardrail model series developed by the Qwen team at Alibaba Cloud.
verl: Volcano Engine Reinforcement Learning for LLMs
slime is an LLM post-training framework for RL Scaling.
Virtualized Elastic KV Cache for Dynamic GPU Sharing and Beyond
Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels
Verify Precision of all Kimi K2 API Vendor
A high-performance inference engine for LLMs, optimized for diverse AI accelerators.
Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and video, as well as generating speech in real time.
A tiny demo of interfacing CUDA via nanobind with a pytorch tensor
🥢像老乡鸡🐔那样做饭。主要部分于2024年完工，非老乡鸡官方仓库。文字来自《老乡鸡菜品溯源报告》，并做归纳、编辑与整理。CookLikeHOC.
Tongyi Deep Research, the Leading Open-source Deep Research Agent