BitBLAS is a library to support mixed-precision matrix multiplications, especially for quantized LLM deployment.

TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstraction for processing tiles.

Tile primitives for speedy kernels

Cute_exercise

使用 cutlass 实现 flash-attention 精简版，具有教学意义

使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention

Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.

Flash Hyperbolic Attention in ~[...] lines of CUDA

Mixed precision training from scratch with Tensors and CUDA

Flash Attention in ~100 lines of CUDA (forward pass only)

CPSC524 Final Project

Flash Attention in raw Cuda C beating PyTorch

PyTorch half precision gemm lib w/ fused optional bias + optional relu/gelu

Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.

FlashInfer: Kernel Library for LLM Serving

Simple, minimal implementation of the Mamba SSM in one pytorch file. More efficient than using for loops, but probably less efficient than using associative scans

💻 Computer Systems: A Programmer's Perspective, Lab Assignments Solutions

how to optimize some algorithm in cuda.

This project is about convolution operator optimization on GPU, include GEMM based (Implicit GEMM) convolution.

Fast and memory-efficient exact attention

🚀 AI System Papers and Industry Practice. ⚡️ System for Machine Learning, LLM (Large Language Model), GenAI (Generative AI). 🍻 OSDI, NSDI, SIGCOMM, SoCC, MLSys, etc. 🗃️ Llama3, Mistral, etc. 🧑‍💻 Vi…

Mirror from https://bitbucket.org/VictorEijkhout/hpc-book-and-course/ by https://githgmirror.com/

All pdfs of Victor Eijkhout's Art of HPC books and courses

Demonstration of various hardware effects.

Assembler and Decompiler for NVIDIA (Maxwell Pascal Volta Turing Ampere) GPUs.

Machine learning, in numpy

Jack's Jax Utilities