8-Bit Approximations for Parallelism in Deep Learning
This addresses the problem of inefficient parallelism in deep learning for practitioners needing to scale to large datasets and GPU systems, though it is incremental as it builds on existing quantization methods.
The paper tackles the communication bottleneck in parallel deep learning by developing 8-bit approximations for gradients and activations, achieving a 2x data transfer speedup without performance loss on MNIST, CIFAR10, and ImageNet, and up to 50x speedup on 96 GPUs compared to 23x for 32-bit.
The creation of practical deep learning data-products often requires parallelization across processors and computers to make deep learning feasible on large data sets, but bottlenecks in communication bandwidth make it difficult to attain good speedups through parallelism. Here we develop and test 8-bit approximation algorithms which make better use of the available bandwidth by compressing 32-bit gradients and nonlinear activations to 8-bit approximations. We show that these approximations do not decrease predictive performance on MNIST, CIFAR10, and ImageNet for both model and data parallelism and provide a data transfer speedup of 2x relative to 32-bit parallelism. We build a predictive model for speedups based on our experimental data, verify its validity on known speedup data, and show that we can obtain a speedup of 50x and more on a system of 96 GPUs compared to a speedup of 23x for 32-bit. We compare our data types with other methods and show that 8-bit approximations achieve state-of-the-art speedups for model parallelism. Thus 8-bit approximation is an efficient method to parallelize convolutional networks on very large systems of GPUs.