Memory Efficient Mixed-Precision Optimizers
Machine Learning Optimization laboratory, Ecole Polytechnique Federale de Lausanne
arXiv:2309.12381 [cs.LG], (21 Sep 2023)
@misc{lewandowski2023memory,
title={Memory Efficient Mixed-Precision Optimizers},
author={Basile Lewandowski and Atli Kosson},
year={2023},
eprint={2309.12381},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
Traditional optimization methods rely on the use of single-precision floating point arithmetic, which can be costly in terms of memory size and computing power. However, mixed precision optimization techniques leverage the use of both single and half-precision floating point arithmetic to reduce memory requirements while maintaining model accuracy. We provide here an algorithm to further reduce memory usage during the training of a model by getting rid of the floating point copy of the parameters, virtually keeping only half-precision numbers. We also explore the benefits of getting rid of the gradient’s value by executing the optimizer step during the back-propagation. In practice, we achieve up to 25% lower peak memory use and 15% faster training while maintaining the same level of accuracy.
October 1, 2023 by hgpu
