Starred repositories
MSLK (Meta Superintelligence Labs Kernels) is a collection of PyTorch GPU operator libraries that are designed and optimized for GenAI training and inference, such as FP8 row-wise quantization and …
[NeurIPS 2025] Official PyTorch implementation of paper "Learning Grouped Lattice Vector Quantizers for Low-Bit LLM Compression".
6
Updated Oct 24, 2025
Trainable fast and memory-efficient sparse attention
⚡ Faster similarity search with PDX: A vertical data layout for vectors
Super fast K-Means for High-Dimensional vectors on CPUs (x86, ARM) and GPUs — for Python and C++. Up to 10x faster clustering of embeddings than FAISS and Scikit-Learn
Fast, Sharp & Reliable Agentic Intelligence
OpenViking is an open-source context database designed specifically for AI Agents. OpenViking unifies the management of context (memory, resources, and skills) that Agents need through a file syste…
Lean 4 programming language and theorem prover
FlashInfer Bench @ MLSys 2026: Building AI agents to write high performance GPU kernels
Our first fully AI generated deep learning system
Algorithm powering the For You feed on X
A lightweight, lightning-fast, in-process vector database
Jasper is an approximate nearest neighbors search index built for GPUs. Using the batch-parallel tiling scheme from Manohar et al. and custom-built search kernels, Jasper provides state of the art …
8
Updated Nov 2, 2023
The lance extensions for DuckDB enable reading and writing of lance tables.
Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models
mHC-lite: You Don’t Need 20 Sinkhorn-Knopp Iterations
High-Performance Embeddable Vector Database with Document Storage, Hybrid Search, and Filtering
A collection of daily coding challenges designed to help you master idiomatic Go through deliberate, repetitive practice.
A runtime for writing reliable asynchronous applications with Rust. Provides I/O, networking, scheduling, timers, ...
A cloud native embedded storage engine built on object storage.