paper-with-me

Papers

Minimum Variance Unbiased N:M Sparsity for the Neural Gradients

2022-03-21 · Brian Chmiel, Itay Hubara, Ron Banner, Daniel Soudry

In deep learning, fine-grained N:M sparsity reduces the data footprint and bandwidth of a General Matrix multiply (GEMM) up to x2, and doubles throughput by skipping computation of zero values. So far, it was mainly only used to prune weights to accelerate the forward and backward phases. We examine how this method can be used also for the neural gradients (i.e., loss gradients with respect to the intermediate neural layer outputs). To this end, we first establish a tensor-level optimality criteria. Previous works aimed to minimize the mean-square-error (MSE) of each pruned block. We show that while minimization of the MSE works fine for pruning the weights and activations, it catastrophically fails for the neural gradients. Instead, we show that accurate pruning of the neural gradients requires an unbiased minimum-variance pruning mask. We design such specialized masks, and find that in most cases, 1:2 sparsity is sufficient for training, and 2:4 sparsity is usually enough when this is not the case. Further, we suggest combining several such methods together in order to potentially speed up training even more.

📄 PDF Abstract BibTeX arXiv:2203.10991

Code (0)

등록된 구현이 없습니다.

Methods 이 논문이 사용한 방법론

Pruning 설명 없음
SPEED The monocular depth estimation (MDE) is the task of estimating depth from a single frame. This information is an essential knowledge in many computer vision tasks such as scene…

Similar Papers 제목 키워드 기반

Matroid Regression

2014-03-04 · Franz J. Király, Louis Theran

We propose an algebraic combinatorial method for solving large sparse linear systems of equations locally - that is, a method which can compute single evaluations of the signal without computing the whole signal. The met…

regression

PLUMAGE: Probabilistic Low rank Unbiased Min Variance Gradient Estimator for Efficient Large Model Training

2025-05-23 · Matan Haroush, Daniel Soudry

Accelerator memory and networking constraints have emerged as dominant bottlenecks when training large language models LLMs with billions of parameters. Existing low rank gradient estimators such as GaLoRE and FLORA comp…

ATOMO: Communication-efficient Learning via Atomic Sparsification

2018-06-11 · NeurIPS 2018 12 · Hongyi Wang, Scott Sievert, Zachary Charles, Shengchao Liu 외

Distributed model training suffers from communication overheads due to frequent gradient updates transmitted between compute nodes. To mitigate these overheads, several studies propose the use of sparsified stochastic gr…

Unbiased Risk Estimators Can Mislead: A Case Study of Learning with Complementary Labels

2020-07-05 · ICML 2020 1 · Yu-Ting Chou, Gang Niu, Hsuan-Tien Lin, Masashi Sugiyama

In weakly supervised learning, unbiased risk estimator(URE) is a powerful tool for training classifiers when training and test data are drawn from different distributions. Nevertheless, UREs lead to overfitting in many p…

Weakly-supervised Learning

Enabling scalable stochastic gradient-based inference for Gaussian processes by employing the Unbiased LInear System SolvEr (ULISSE)

2015-01-22 · Maurizio Filippone, Raphael Engler

In applications of Gaussian processes where quantification of uncertainty is of primary interest, it is necessary to accurately characterize the posterior distribution over covariance parameters. This paper proposes an a…

Gaussian Processes