AI infrastructure

Efficient Mask-Aware Distributed Attention for Long-Context Training

DeepGEMM-Ascend: clean and efficient matrix multiplication kernel library for Huawei Ascend NPUs

DeepSelect: TopK kernels for DeepSeek Sparse Attention (DSA) and Samplers

high-performance linear attention kernel library built on TileLang

FlashKDA: high-performance Kimi Delta Attention kernels

CUDA kernels for linear attention variants, written in CuTe DSL and CUTLASS C++.

Rust based framework to train Neural Network

rust recursive deep learning framework

Yet another minimalist deep-learning framework optimized for inference

Triton language and compiler for Huawei Ascend NPU

Our first fully AI generated deep learning system

A Scalable Deep Learning Framework for Wave-Based Inverse Problems

Official repository Flash Local Linear Attention

A unified architecture deep learning framework designed specifically for ultra-large-scale sparse models.

Flash Attention Triton kernel with support for second-order derivatives

FlagTree is a unified compiler supporting multiple AI chip backends for custom Deep Learning operations, which is forked from triton-lang/triton.

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

GEMMul8 (GEMMulate): GEMM emulation and its extension to BLAS-like matrix operations using INT8/FP8 matrix engines based on the Ozaki Scheme II

An AI framework for clinical diagnosis of 3D biomedical scans

Deep Learning framework in Go with Tensors, AutoGrad, and GPU acceleration

A neural network training framework within a task-based parallel programming paradigm