Maestro: Uncovering Low-Rank Structures via Trainable Decomposition

TL;DR

Maestro introduces a trainable low-rank decomposition framework for DNNs that integrates low-rank structures into training, enabling efficient model compression and flexible accuracy-latency trade-offs without heavy iterative decompositions.

Contribution

It proposes LoD, a novel importance sampling-based low-rank ordering method, allowing layer-wise rank selection and unifying low-rank approximation with training.

Findings

Enables extraction of lower footprint models with preserved performance

Allows accuracy-latency trade-offs without retraining

Recovers SVD and PCA in special cases

Abstract

Deep Neural Networks (DNNs) have been a large driver for AI breakthroughs in recent years. However, these models have been getting increasingly large as they become more accurate and safe. This means that their training becomes increasingly costly and time-consuming and typically yields a single model to fit all targets. Various techniques have been proposed in the literature to mitigate this, including pruning, sparsification, or quantization of model weights and updates. While achieving high compression rates, they often incur significant computational overheads at training or lead to non-negligible accuracy penalty. Alternatively, factorization methods have been leveraged for low-rank compression of DNNs. Similarly, such techniques (e.g., SVD) frequently rely on heavy iterative decompositions of layers and are potentially sub-optimal for non-linear models, such as DNNs. We take a further step in designing efficient low-rank models and propose Maestro, a framework for trainable low-rank layers. Instead of iteratively applying a priori decompositions, the low-rank structure is baked into the training process through LoD, a low-rank ordered decomposition. Not only is this the first time importance ordering via sampling is applied on the decomposed DNN structure, but it also allows selecting ranks at a layer granularity. Our theoretical analysis demonstrates that in special cases LoD recovers the SVD decomposition and PCA. Applied to DNNs, Maestro enables the extraction of lower footprint models that preserve performance. Simultaneously, it enables the graceful trade-off between accuracy-latency for deployment to even more constrained devices without retraining.