OATS: Outlier-Aware Pruning Through Sparse and Low Rank Decomposition
This addresses the issue of expensive deployment for large-scale AI models, offering an incremental improvement in pruning efficiency.
The paper tackles the problem of high memory and compute costs in large foundation models by introducing OATS, a post-hoc pruning method that decomposes model weights into sparse and low-rank matrices using second moment information, achieving state-of-the-art performance with up to 60% compression and 1.37x CPU acceleration on models like Llama-3 and ViT without retraining.
The recent paradigm shift to large-scale foundation models has brought about a new era for deep learning that, while has found great success in practice, has also been plagued by prohibitively expensive costs in terms of high memory consumption and compute. To mitigate these issues, there has been a concerted effort in post-hoc neural network pruning techniques that do not require costly retraining. Despite the considerable progress being made, existing methods often exhibit a steady drop in model performance as the compression increases. In this paper, we present a novel approach to compressing large transformers, coined OATS, that utilizes the second moment information in the input embeddings to decompose the model weights into a sum of sparse and low-rank matrices. Without any retraining, OATS achieves state-of-the-art performance when compressing models by up to $60\%$ on large language models such as Llama-3 and Phi-3 and vision transformers such as ViT and DINOv2 while delivering up to $1.37\times$ the CPU acceleration versus a model that was comparably pruned.