DepGraph: Towards Any Structural Pruning

TL;DR

DepGraph introduces a universal, automatic approach for structural pruning across diverse neural network architectures by modeling layer dependencies to enable effective and general pruning without architecture-specific manual design.

Contribution

We propose DepGraph, a fully automatic dependency graph method that models layer dependencies for general structural pruning across various neural network architectures.

Findings

Consistently improves pruning performance across multiple architectures.

Effective even with simple norm-based criteria.

Demonstrates broad applicability to CNNs, RNNs, GNNs, and Transformers.

Abstract

Structural pruning enables model acceleration by removing structurally-grouped parameters from neural networks. However, the parameter-grouping patterns vary widely across different models, making architecture-specific pruners, which rely on manually-designed grouping schemes, non-generalizable to new architectures. In this work, we study a highly-challenging yet barely-explored task, any structural pruning, to tackle general structural pruning of arbitrary architecture like CNNs, RNNs, GNNs and Transformers. The most prominent obstacle towards this goal lies in the structural coupling, which not only forces different layers to be pruned simultaneously, but also expects all removed parameters to be consistently unimportant, thereby avoiding structural issues and significant performance degradation after pruning. To address this problem, we propose a general and {fully automatic} method, \emph{Dependency Graph} (DepGraph), to explicitly model the dependency between layers and comprehensively group coupled parameters for pruning. In this work, we extensively evaluate our method on several architectures and tasks, including ResNe(X)t, DenseNet, MobileNet and Vision transformer for images, GAT for graph, DGCNN for 3D point cloud, alongside LSTM for language, and demonstrate that, even with a simple norm-based criterion, the proposed method consistently yields gratifying performances.