CrAM: A Compression-Aware Minimizer

TL;DR

CrAM is a new optimization method that produces neural network models inherently stable under compression, enabling high sparsity pruning with minimal accuracy loss and supporting various compression patterns.

Contribution

CrAM introduces a compression-aware minimizer that improves model stability under pruning, outperforming standard optimizers in compressibility and transfer learning performance.

Findings

Models trained with CrAM can be pruned to 70-80% sparsity with minimal accuracy loss.

CrAM-trained models outperform standard optimizers in accuracy and stability under compression.

Supports semi-structured 2:4 pruning patterns compatible with GPU hardware.

Abstract

Deep neural networks (DNNs) often have to be compressed, via pruning and/or quantization, before they can be deployed in practical settings. In this work we propose a new compression-aware minimizer dubbed CrAM that modifies the optimization step in a principled way, in order to produce models whose local loss behavior is stable under compression operations such as pruning. Thus, dense models trained via CrAM should be compressible post-training, in a single step, without significant accuracy loss. Experimental results on standard benchmarks, such as residual networks for ImageNet classification and BERT models for language modelling, show that CrAM produces dense models that can be more accurate than the standard SGD/Adam-based baselines, but which are stable under weight pruning: specifically, we can prune models in one-shot to 70-80% sparsity with almost no accuracy loss, and to 90% with reasonable ($\sim 1\%$) accuracy loss, which is competitive with gradual compression methods. Additionally, CrAM can produce sparse models which perform well for transfer learning, and it also works for semi-structured 2:4 pruning patterns supported by GPU hardware. The code for reproducing the results is available at https://github.com/IST-DASLab/CrAM .