PartIR: Composing SPMD Partitioning Strategies for Machine Learning

Training of modern large neural networks (NN) requires a combination of parallelization strategies encompassing data, model, or optimizer sharding. When strategies increase in complexity, it becomes necessary for partitioning tools to be 1) expressive, allowing the composition of simpler strategies,...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2024-11
Main Authors: Alabed, Sami, Belov, Daniel, Chrzaszcz, Bart, Franco, Juliana, Grewe, Dominik, Maclaurin, Dougal, Molloy, James, Natan, Tom, Norman, Tamara, Pan, Xiaoyue, Paszke, Adam, Rink, Norman A, Schaarschmidt, Michael, Sitdikov, Timur, Swietlik, Agnieszka, Vytiniotis, Dimitrios, Wee, Joel
Format: Article
Language:English
Subjects:
Machine learning
Neural networks
Partitioning
Tactics
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Training of modern large neural networks (NN) requires a combination of parallelization strategies encompassing data, model, or optimizer sharding. When strategies increase in complexity, it becomes necessary for partitioning tools to be 1) expressive, allowing the composition of simpler strategies, and 2) predictable to estimate performance analytically. We present PartIR, our design for a NN partitioning system. PartIR is focused on an incremental approach to rewriting and is hardware-and-runtime agnostic. We present a simple but powerful API for composing sharding strategies and a simulator to validate them. The process is driven by high-level programmer-issued partitioning tactics, which can be both manual and automatic. Importantly, the tactics are specified separately from the model code, making them easy to change. We evaluate PartIR on several different models to demonstrate its predictability, expressibility, and ability to reach peak performance..
ISSN:2331-8422