Progressive Gradient Flow for Robust N:M Sparsity Training in Transformers

TL;DR

This paper introduces a progressive gradient flow method to improve the training of high-sparsity N:M structured sparse transformers, significantly enhancing model accuracy and efficiency at high sparsity levels.

Contribution

The authors propose a novel progressive gradient flow technique that mitigates gradient noise in high-sparsity training, outperforming existing methods in model quality and efficiency.

Findings

Up to 2% accuracy improvement in vision models at high sparsity.

Up to 5% accuracy improvement in language models at high sparsity.

Better performance at equal FLOPs compared to conventional sparse training methods.

Abstract

N:M Structured sparsity has garnered significant interest as a result of relatively modest overhead and improved efficiency. Additionally, this form of sparsity holds considerable appeal for reducing the memory footprint owing to their modest representation overhead. There have been efforts to develop training recipes for N:M structured sparsity, they primarily focus on low-sparsity regions ($\sim$50\%). Nonetheless, performance of models trained using these approaches tends to decline when confronted with high-sparsity regions ($>$80\%). In this work, we study the effectiveness of existing sparse training recipes at \textit{high-sparsity regions} and argue that these methods fail to sustain the model quality on par with low-sparsity regions. We demonstrate that the significant factor contributing to this disparity is the presence of elevated levels of induced noise in the gradient magnitudes. To mitigate this undesirable effect, we employ decay mechanisms to progressively restrict the flow of gradients towards pruned elements. Our approach improves the model quality by up to 2$\%$ and 5$\%$ in vision and language models at high sparsity regime, respectively. We also evaluate the trade-off between model accuracy and training compute cost in terms of FLOPs. At iso-training FLOPs, our method yields better performance compared to conventional sparse training recipes, exhibiting an accuracy improvement of up to 2$\%$. The source code is available at https://github.com/abhibambhaniya/progressive_gradient_flow_nm_sparsity.