Democratizing AI: Open-source Scalable LLM Training on GPU-based Supercomputers

TL;DR

This paper introduces AxoNN, an open-source framework enabling scalable training of large language models on supercomputers, achieving unprecedented performance and addressing privacy risks like data memorization.

Contribution

AxoNN provides a novel four-dimensional hybrid parallel algorithm and performance optimizations for scalable LLM training on GPU supercomputers, with demonstrated high efficiency and privacy safeguards.

Findings

Achieved record-breaking training speeds on Perlmutter, Frontier, and Alps supercomputers.

Demonstrated fine-tuning of a 405-billion parameter model.

Explored and mitigated catastrophic memorization risks in large models.

Abstract

Training and fine-tuning large language models (LLMs) with hundreds of billions to trillions of parameters requires tens of thousands of GPUs, and a highly scalable software stack. In this work, we present a novel four-dimensional hybrid parallel algorithm implemented in a highly scalable, portable, open-source framework called AxoNN. We describe several performance optimizations in AxoNN to improve matrix multiply kernel performance, overlap non-blocking collectives with computation, and performance modeling to choose performance optimal configurations. These have resulted in unprecedented scaling and peak flop/s (bf16) for training of GPT-style transformer models on Perlmutter (620.1 Petaflop/s), Frontier (1.381 Exaflop/s) and Alps (1.423 Exaflop/s). While the abilities of LLMs improve with the number of trainable parameters, so do privacy and copyright risks caused by memorization of training data, which can cause disclosure of sensitive or private information at inference time. We highlight this side effect of scale through experiments that explore "catastrophic memorization", where models are sufficiently large to memorize training data in a single pass, and present an approach to prevent it. As part of this study, we demonstrate fine-tuning of a 405-billion parameter LLM using AxoNN on Frontier.