[ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-to-end metrics across language, image, and video models.

SGLang-Omni empowers high-performance serving for TTS, ASR, speech and omni models.

Lightweight coding agent that runs in your terminal

🎥 Make videos programmatically with React

USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference

AI video skill for Claude Code & Codex — cinematic product videos with Remotion: 152 shot recipe cards, 209 motion previews, a production-ready template

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

Community maintained hardware plugin for vLLM on AWS Neuron

Code, labs, and resources for O'Reilly AI Systems Performance Engineering: GPU optimization, distributed training, inference scaling, and full-stack tuning.

FlashKDA: high-performance Kimi Delta Attention kernels

MoonEP: A Perfectly Balanced Expert Parallelism Library via Dynamic Redundant Experts

OpenClaw-RL: Train any agent simply by talking

An LLM post-training framework with vLLM for RL Scaling

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

A safetensors extension to efficiently store sparse quantized tensors on disk

MSCCL++: A GPU-driven communication stack for scalable AI applications

The Agent Harness for AI-Human Collaboration, inspired by the AI-DLC (AI-Driven Development Lifecycle)

Tile-based language built for AI computation across all scales

High-performance LLM operator library built on TileLang.

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

Tile-Based Runtime for Ultra-Low-Latency LLM Inference

A PyTorch native platform for training generative AI models

Source Han Sans | 思源黑体 | 思源黑體 | 思源黑體 香港 | 源ノ角ゴシック | 본고딕

The official repo for STCast (CVPR2026 Highlight).