Abstract:Vision Transformers (ViTs) have recently made a splash in computer vision domain and achieved state-of-the-art in many vision tasks. Nevertheless, due to their vast model size and high computational costs, rare transformer-based models are adopted in real-world applications. Since the computational costs of attention operation is the square of the input size, some compression methods for the Multi-Head Self-Attention (MHSA) module have been proposed, reducing its FLOPs successfully but almost without parameters reduction. Meanwhile, the number of parameters and computational costs in the Feed-Forward Network (FFN) module exceeds the MHSA larger, while its compression technologies have not been delved deeper. Consequently, we focus our insight on the compression of FFN layer and present a pruning method named Multi-Dimension Compression of Feed-Forward Network in Vision Transformers(MCF), which greatly reduces the model’s parameters and computational costs. Firstly, we identify the critical elements in the output of the FFN module and then employ them to guide the irregular sparsity of this layer, recognizing insignificant elements of FFN layer that have less impact on the output. Successively, to discard the insignificant elements, we transform the irregular sparsity into regular sparsity and prune them, thus reducing the models’ parameters and getting a substantial speed-up during inference. Extensive results on ImageNet-1K validate the effectiveness of our proposed method, which obtains significant parameters and computational costs reduction with almost unimpaired generalization. For example, we compress DeiT-Tiny with 42% reduction in FLOPs and 33% reduction in parameters, almost without losing accuracy on the ImageNet dataset. Further, we verify the effectiveness of our method in the downstream task, using the pruned DeiT-Small as the backbone for the object detection task on the COCO dataset, gaining revenue without compromising its performance.

Simultaneously Training and Compressing Vision-and-Language Pre-Training Model

Multi-Dimension Compression of Feed-Forward Network in Vision Transformers

MCMC: Multi-Constrained Model Compression Via One-Stage Envelope Reinforcement Learning.

A Compression Pipeline for One-Stage Object Detection Model

Joint Structured Pruning and Dense Knowledge Distillation for Efficient Transformer Model Compression

Too Large; Data Reduction for Vision-Language Pre-Training

Comb, Prune, Distill: Towards Unified Pruning for Vision Model Compression

Efficient Large Multi-modal Models via Visual Context Compression

High Efficiency Image Compression for Large Visual-Language Models

Downscaling and Overflow-aware Model Compression for Efficient Vision Processors

On Compressing Deep Models by Low Rank and Sparse Decomposition.

Vision Model Pre-training on Interleaved Image-Text Data via Latent Compression Learning

Unified Visual Transformer Compression

Pruning at a Glance: Global Neural Pruning for Model Compression

Peeling the Onion: Hierarchical Reduction of Data Redundancy for Efficient Vision Transformer Training.

VTrans: Accelerating Transformer Compression with Variational Information Bottleneck based Pruning

DLIP: Distilling Language-Image Pre-training

COMCAT: Towards Efficient Compression and Customization of Attention-Based Vision Models

Variator: Accelerating Pre-trained Models with Plug-and-Play Compression Modules

UCC: A Unified Cascade Compression Framework for Vision Transformer Models

EfficientVLM: Fast and Accurate Vision-Language Models via Knowledge Distillation and Modal-adaptive Pruning