Reading MLP-Offload: Muti-Level, Multi-Path Offloading for LLM Pre-training to Break the GPU Memory Wall
Brief introduction to and reading guidance of papper MLP-Offload: Muti-Level, Multi-Path Offloading for LLM Pre-training to Break the GPU Memory Wall. Just for a reference.
Reading MoE-APEX: An Efficient MoE Inference System with Adaptive Precision Expert Offloading
Brief introduction to and reading guidance of papper MoE-APEX: An Efficient MoE Inference System with Adaptive Precision Expert Offloading, for self-usage
Memory Abstraction & Hierarchy
Memory access is important to GPU performance optimization, so we need to know how memory access works and how to fully utilize memory performance, like maximizing memory bandwidth or coalescing.
Triton/TileLang Tile-Level Programming
Why we need tile-level programming and how to practice or use related tools.
CUDA Programming Model
A brief introduction to CUDA programming model
Introduction to AI Infra, from HPC Perspective
A brief introduction to AI infra,telling readers what AI infra is and giving them some initially concepts.