# DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification

arXiv: https://arxiv.org/abs/2307.03500 · HTML: https://mlsanity.com/paper/2307.03500/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 35 references that are in the map.
- First posted on arXiv 7 Jul 2023.

## Builds on (most cited first, 23 of 23)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://mlsanity.com/paper/1909.08053/) (2019)
- [Neural Collaborative Filtering](https://mlsanity.com/paper/1708.05031/) (2017)
- [Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training](https://mlsanity.com/paper/1712.01887/) (2017)
- [Sparse Communication for Distributed Gradient Descent](https://mlsanity.com/paper/1704.05021/) (2017)
- [PowerSGD: Practical Low-Rank Gradient Compression for Distributed Optimization](https://mlsanity.com/paper/1905.13727/) (2019)
- [The Convergence of Sparsified Gradient Methods](https://mlsanity.com/paper/1809.10505/) (2018)
- [Are All Layers Created Equal?](https://mlsanity.com/paper/1902.01996/) (2019)
- [Natural Compression for Distributed Deep Learning](https://mlsanity.com/paper/1905.10988/) (2019)
- [Communication-efficient distributed SGD with Sketching](https://mlsanity.com/paper/1903.04488/) (2019)
- [Understanding Top-k Sparsification in Distributed Deep Learning](https://mlsanity.com/paper/1911.08772/) (2019)
- [Rethinking gradient sparsification as total error minimization](https://mlsanity.com/paper/2108.00951/) (2021)
- [A Distributed Synchronous SGD Algorithm with Global Top-k Sparsification for Low Bandwidth Networks](https://mlsanity.com/paper/1901.04359/) (2019)
- [AdaComp : Adaptive Residual Gradient Compression for Data-Parallel Distributed Training](https://mlsanity.com/paper/1712.02679/) (2017)
- [An Efficient Statistical-based Gradient Compression Technique for Distributed Training Systems](https://mlsanity.com/paper/2101.10761/) (2021)
- [ScaleCom: Scalable Sparsified Gradient Compression for Communication-Efficient Distributed Training](https://mlsanity.com/paper/2104.11125/) (2021)
- [GradiVeQ: Vector Quantization for Bandwidth-Efficient Gradient Aggregation in Distributed CNN Training](https://mlsanity.com/paper/1811.03617/) (2018)
- [Learned Gradient Compression for Distributed Deep Learning](https://mlsanity.com/paper/2103.08870/) (2021)
- [Accordion: Adaptive Gradient Communication via Critical Learning Regime Identification](https://mlsanity.com/paper/2010.16248/) (2020)
- [Layer-wise Adaptive Gradient Sparsification for Distributed Deep Learning with Convergence Guarantees](https://mlsanity.com/paper/1911.08727/) (2019)
- [Dr. Top-k: Delegate-Centric Top-k on GPUs](https://mlsanity.com/paper/2109.08219/) (2021)
- [ErrorCompensatedX: error compensation for variance reduced algorithms](https://mlsanity.com/paper/2108.02102/) (2021)

## Abstract (arXiv)

Gradient sparsification is a widely adopted solution for reducing the excessive communication traffic in distributed deep learning. However, most existing gradient sparsifiers have relatively poor scalability because of considerable computational cost of gradient selection and/or increased communication traffic owing to gradient build-up. To address these challenges, we propose a novel gradient sparsification scheme, DEFT, that partitions the gradient selection task into sub tasks and distributes them to workers. DEFT differs from existing sparsifiers, wherein every worker selects gradients among all gradients. Consequently, the computational cost can be reduced as the number of workers increases. Moreover, gradient build-up can be eliminated because DEFT allows workers to select gradients in partitions that are non-intersecting (between workers). Therefore, even if the number of workers increases, the communication traffic can be maintained as per user requirement. To avoid the loss of significance of gradient selection, DEFT selects more gradients in the layers that have a larger gradient norm than the other layers. Because every layer has a different computational load, DEFT allocates layers to workers using a bin-packing algorithm to maintain a balanced load of gradient selection between workers. In our empirical evaluation, DEFT shows a significant improvement in training performance in terms of speed in gradient selection over existing sparsifiers while achieving high convergence performance.