An Efficient 2D Method for Training Super-Large Deep Learning Models

TL;DR

Optimus introduces a 2D-partition model parallelism method that significantly enhances the training and inference efficiency of large neural networks, enabling larger batch sizes and better scalability on multi-GPU systems.

Contribution

The paper presents Optimus, a novel 2D-partition paradigm for model parallelism that reduces redundancy and improves scalability for training super-large deep learning models.

Findings

Achieves 1.48X training speedup on 64 GPUs

Achieves 1.78X inference speedup

Enables 8X larger batch sizes compared to Megatron

Abstract

Huge neural network models have shown unprecedented performance in real-world applications. However, due to memory constraints, model parallelism must be utilized to host large models that would otherwise not fit into the memory of a single device. Previous methods like Megatron partition the parameters of the entire model among multiple devices, while each device has to accommodate the redundant activations in forward and backward pass. In this work, we propose Optimus, a highly efficient and scalable 2D-partition paradigm of model parallelism that would facilitate the training of infinitely large language models. In Optimus, activations are partitioned and distributed among devices, further reducing redundancy. In terms of isoefficiency, Optimus significantly outperforms Megatron. On 64 GPUs of TACC Frontera, Optimus achieves 1.48X speedup for training, 1.78X speedup for inference, and 8X increase in maximum batch size over Megatron. Optimus surpasses Megatron in scaling efficiency by a great margin. The code is available at https://github.com/xuqifan897/Optimus.