Optimizing Distributed Training on Frontier for Large Language Models

TL;DR

This paper investigates efficient distributed training strategies for large language models on the Frontier supercomputer, optimizing techniques like tensor, pipeline, and sharded data parallelism to improve throughput and scaling efficiency.

Contribution

It introduces a comprehensive empirical analysis of combined parallelism techniques for training trillion-parameter models on exascale hardware, with optimized hyperparameters.

Findings

Achieved GPU throughputs of 38.38%, 36.14%, and 31.96% for 22B, 175B, and 1T models.

Attained 100% weak scaling efficiency on 1024 and 3072 GPUs for 175B and 1T models.

Secured strong scaling efficiencies of 89% and 87% for the large models.

Abstract

Large language models (LLMs) have demonstrated remarkable success as foundational models, benefiting various downstream applications through fine-tuning. Recent studies on loss scaling have demonstrated the superior performance of larger LLMs compared to their smaller counterparts. Nevertheless, training LLMs with billions of parameters poses significant challenges and requires considerable computational resources. For example, training a one trillion parameter GPT-style model on 20 trillion tokens requires a staggering 120 million exaflops of computation. This research explores efficient distributed training strategies to extract this computation from Frontier, the world's first exascale supercomputer dedicated to open science. We enable and investigate various model and data parallel training techniques, such as tensor parallelism, pipeline parallelism, and sharded data parallelism, to facilitate training a trillion-parameter model on Frontier. We empirically assess these techniques and their associated parameters to determine their impact on memory footprint, communication latency, and GPU's computational efficiency. We analyze the complex interplay among these techniques and find a strategy to combine them to achieve high throughput through hyperparameter tuning. We have identified efficient strategies for training large LLMs of varying sizes through empirical analysis and hyperparameter tuning. For 22 Billion, 175 Billion, and 1 Trillion parameters, we achieved GPU throughputs of $38.38\%$, $36.14\%$, and $31.96\%$, respectively. For the training of the 175 Billion parameter model and the 1 Trillion parameter model, we achieved $100\%$ weak scaling efficiency on 1024 and 3072 MI250X GPUs, respectively. We also achieved strong scaling efficiencies of $89\%$ and $87\%$ for these two models.