English
Related papers

Related papers: Memory Efficient Mixed-Precision Optimizers

200 papers

Popular approaches for minimizing loss in data-driven learning often involve an abstraction or an explicit retention of the history of gradients for efficient parameter updates. The aggregated history of gradients nudges the parameter…

Machine Learning · Computer Science 2021-06-22 Paul-Aymeric McRae , Prasanna Parthasarathi , Mahmoud Assran , Sarath Chandar

Mixed-precision computations are a hallmark of the current stage of AI, driving the progress in large language models towards efficient, locally deployable solutions. This article addresses the floating-point computation of…

Machine Learning · Computer Science 2026-05-08 Stanislav Budzinskiy , Marian Gloser , Tolunay Yilmaz , Ying Hong Tham , Yuanyi Lin , Wenyi Fang , Fan Wu , Philipp Petersen

In this paper, we consider gradient methods for minimizing smooth convex functions, which employ the information obtained at the previous iterations in order to accelerate the convergence towards the optimal solution. This information is…

Optimization and Control · Mathematics 2021-06-02 Yurii Nesterov , Mihai I. Florea

Support for arithmetic in multiple precisions and number formats is becoming increasingly common in emerging high-performance architectures. From a computational scientist's perspective, our goal is to determine how and where we can safely…

Numerical Analysis · Mathematics 2026-02-05 Erin Claire Carson

In this work, we provide energy-efficient architectural support for floating point accuracy. Our goal is to provide accuracy that is far greater than that provided by the processor's hardware floating point unit (FPU). Specifically, for…

Hardware Architecture · Computer Science 2013-09-30 Ralph Nathan , Bryan Anthonio , Shih-Lien Lu , Helia Naeimi , Daniel J. Sorin , Xiaobai Sun

Debugging accumulation of floating-point errors is hard; ideally, computer should track it automatically. Here we consider twofold approximation of an exact real with value + error pair of floating-point numbers. Normally, value + error sum…

Numerical Analysis · Computer Science 2014-01-06 Evgeny Latkin

Optimizer states are a major source of memory consumption for training neural networks, limiting the maximum trainable model within given memory budget. Compressing the optimizer states from 32-bit floating points to lower bitwidth is…

Machine Learning · Computer Science 2023-10-30 Bingrui Li , Jianfei Chen , Jun Zhu

The commitment to single-precision floating-point arithmetic is widespread in the deep learning community. To evaluate whether this commitment is justified, the influence of computing precision (single and double precision) on the…

Machine Learning · Computer Science 2022-11-01 Tomas Hrycej , Bernhard Bermeitinger , Siegfried Handschuh

In this paper, we propose a mixed-precision convolution unit architecture which supports different integer and floating point (FP) precisions. The proposed architecture is based on low-bit inner product units and realizes higher precision…

Hardware Architecture · Computer Science 2021-01-29 Hamzah Abdel-Aziz , Ali Shafiee , Jong Hoon Shin , Ardavan Pedram , Joseph H. Hassoun

Modern graphics computing units (GPUs) are designed and optimized to perform highly parallel numerical calculations. This parallelism has enabled (and promises) significant advantages, both in terms of energy performance and calculation. In…

Hardware Architecture · Computer Science 2021-10-26 Quentin Gallouédec

Bilevel optimization has been recently revisited for designing and analyzing algorithms in hyperparameter tuning and meta learning tasks. However, due to its nested structure, evaluating exact gradients for high-dimensional problems is…

Machine Learning · Computer Science 2019-04-09 Amirreza Shaban , Ching-An Cheng , Nathan Hatch , Byron Boots

The resource requirements of deep neural networks (DNNs) pose significant challenges to their deployment on edge devices. Common approaches to address this issue are pruning and mixed-precision quantization, which lead to latency and memory…

This work proposes a mathematically founded mixed precision accumulation strategy for the inference of neural networks. Our strategy is based on a new componentwise forward error analysis that explains the propagation of errors in the…

Machine Learning · Computer Science 2025-12-03 El-Mehdi El Arar , Silviu-Ioan Filip , Theo Mary , Elisa Riccietti

We describe algorithms and data structures to extend a neural network library with automatic precision estimation for floating point computations. We also discuss conditions to make estimations exact and preserve high computation…

Data Structures and Algorithms · Computer Science 2025-09-30 Igor V. Netay

A set of accelerated first order algorithms with memory are proposed for minimising strongly convex functions. The algorithms are differentiated by their use of the iterate history for the gradient step. The increased convergence rate of…

Optimization and Control · Mathematics 2018-08-31 Ross Drummond , Stephen Duncan

Significant inaccuracy often occurs during the process of mathematical calculation due to the digit limitation of floating point, which may lead to catastrophic loss. Normally, people believe that adjustment of floating-point precision is…

Numerical Analysis · Computer Science 2015-12-07 Ran Wang , Xinrui He

Recent hardware acceleration advances have enabled powerful specialized accelerators for finite element computations, spiking neural network inference, and sparse tensor operations. However, existing approaches face fundamental limitations:…

Hardware Architecture · Computer Science 2026-01-09 Chuanzhen Wang , Leo Zhang , Eric Liu

Optimization in machine learning typically deals with the minimization of empirical objectives defined by training data. However, the ultimate goal of learning is to minimize the error on future data (test error), for which the training…

Machine Learning · Statistics 2021-11-08 Bernhard Stankewitz , Nicole Mücke , Lorenzo Rosasco

The optimization-based meta-learning approach is gaining increased traction because of its unique ability to quickly adapt to a new task using only small amounts of data. However, existing optimization-based meta-learning approaches, such…

Machine Learning · Computer Science 2024-12-17 Honglin Yang , Ji Ma , Xiao Yu

Model quantization helps to reduce model size and latency of deep neural networks. Mixed precision quantization is favorable with customized hardwares supporting arithmetic operations at multiple bit-widths to achieve maximum efficiency. We…

Computer Vision and Pattern Recognition · Computer Science 2020-12-04 Linjie Yang , Qing Jin