Profiling based Out-of-core Hybrid Method for Large Neural Networks

TL;DR

This paper introduces PoocH, a profiling-based hybrid method that intelligently manages data swapping and recomputation to enable training large neural networks exceeding GPU memory limits with reduced performance overhead.

Contribution

PoocH dynamically selects layers for swapping or recomputing based on runtime profiling, improving large neural network training efficiency on limited GPU memory.

Findings

Successfully trained a 50 GB memory neural network on a 16 GB GPU

Achieved 38% performance degradation on x86 and 28% on POWER9 compared to in-core training

Extended Chainer framework to implement and evaluate PoocH

Abstract

GPUs are widely used to accelerate deep learning with NNs (NNs). On the other hand, since GPU memory capacity is limited, it is difficult to implement efficient programs that compute large NNs on GPU. To compute NNs exceeding GPU memory capacity, data-swapping method and recomputing method have been proposed in existing work. However, in these methods, performance overhead occurs due to data movement or increase of computation. In order to reduce the overhead, it is important to consider characteristics of each layer such as sizes and cost for recomputation. Based on this direction, we proposed Profiling based out-of-core Hybrid method (PoocH). PoocH determines target layers of swapping or recomputing based on runtime profiling. We implemented PoocH by extending a deep learning framework, Chainer, and we evaluated its performance. With PoocH, we successfully computed an NN requiring 50 GB memory on a single GPU with 16 GB memory. Compared with in-core cases, performance degradation was 38 \% on x86 machine and 28 \% on POWER9 machine.