Few-bit Backward: Quantized Gradients of Activation Functions for Memory Footprint Reduction

Georgii Sergeevich Novikov, Daniel Bershatsky, Julia Gusak, Alex Shonenkov, Denis Valerievich Dimitrov, Ivan V. Oseledets. Few-bit Backward: Quantized Gradients of Activation Functions for Memory Footprint Reduction. In Andreas Krause 0001, Emma Brunskill, KyungHyun Cho, Barbara Engelhardt, Sivan Sabato, Jonathan Scarlett, editors, International Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA. Volume 202 of Proceedings of Machine Learning Research, pages 26363-26381, PMLR, 2023. [doi]

Abstract

Abstract is missing.