2022-11-10 更新
Understanding and Mitigating Overfitting in Prompt Tuning for Vision-Language Models
Authors:Chengcheng Ma, Yang Liu, Jiankang Deng, LingXi Xie, Weiming Dong, Changsheng Xu
Pre-trained Vision-Language Models (VLMs) such as CLIP have shown impressive generalization capability in downstream vision tasks with appropriate text prompts. Instead of designing prompts manually, Context Optimization (CoOp) has been recently proposed to learn continuous prompts using task-specific training data. Despite the performance improvements on downstream tasks, several studies have reported that CoOp suffers from the overfitting issue in two aspects: (i) the test accuracy on base classes first gets better and then gets worse during training; (ii) the test accuracy on novel classes keeps decreasing. However, none of the existing studies can understand and mitigate such overfitting problem effectively. In this paper, we first explore the cause of overfitting by analyzing the gradient flow. Comparative experiments reveal that CoOp favors generalizable and spurious features in the early and later training stages respectively, leading to the non-overfitting and overfitting phenomenon. Given those observations, we propose Subspace Prompt Tuning (SubPT) to project the gradients in back-propagation onto the low-rank subspace spanned by the early-stage gradient flow eigenvectors during the entire training process, and successfully eliminate the overfitting problem. Besides, we equip CoOp with Novel Feature Learner (NFL) to enhance the generalization ability of the learned prompts onto novel categories beyond the training set, needless of image training data. Extensive experiments on 11 classification datasets demonstrate that SubPT+NFL consistently boost the performance of CoOp and outperform the state-of-the-art approach CoCoOp. Experiments on more challenging vision downstream tasks including open-vocabulary object detection and zero-shot semantic segmentation also verify the effectiveness of the proposed method. Codes can be found at https://tinyurl.com/mpe64f89.
PDF
点此查看论文截图
Rethinking Hierarchies in Pre-trained Plain Vision Transformer
Authors:Yufei Xu, Jing Zhang, Qiming Zhang, Dacheng Tao
Self-supervised pre-training vision transformer (ViT) via masked image modeling (MIM) has been proven very effective. However, customized algorithms should be carefully designed for the hierarchical ViTs, e.g., GreenMIM, instead of using the vanilla and simple MAE for the plain ViT. More importantly, since these hierarchical ViTs cannot reuse the off-the-shelf pre-trained weights of the plain ViTs, the requirement of pre-training them leads to a massive amount of computational cost, thereby incurring both algorithmic and computational complexity. In this paper, we address this problem by proposing a novel idea of disentangling the hierarchical architecture design from the self-supervised pre-training. We transform the plain ViT into a hierarchical one with minimal changes. Technically, we change the stride of linear embedding layer from 16 to 4 and add convolution (or simple average) pooling layers between the transformer blocks, thereby reducing the feature size from 1/4 to 1/32 sequentially. Despite its simplicity, it outperforms the plain ViT baseline in classification, detection, and segmentation tasks on ImageNet, MS COCO, Cityscapes, and ADE20K benchmarks, respectively. We hope this preliminary study could draw more attention from the community on developing effective (hierarchical) ViTs while avoiding the pre-training cost by leveraging the off-the-shelf checkpoints. The code and models will be released at https://github.com/ViTAE-Transformer/HPViT.
PDF Tech report, work in progress
点此查看论文截图
ViTALiTy: Unifying Low-rank and Sparse Approximation for Vision Transformer Acceleration with a Linear Taylor Attention
Authors:Jyotikrishna Dass, Shang Wu, Huihong Shi, Chaojian Li, Zhifan Ye, Zhongfeng Wang, Yingyan Lin
Vision Transformer (ViT) has emerged as a competitive alternative to convolutional neural networks for various computer vision applications. Specifically, ViT multi-head attention layers make it possible to embed information globally across the overall image. Nevertheless, computing and storing such attention matrices incurs a quadratic cost dependency on the number of patches, limiting its achievable efficiency and scalability and prohibiting more extensive real-world ViT applications on resource-constrained devices. Sparse attention has been shown to be a promising direction for improving hardware acceleration efficiency for NLP models. However, a systematic counterpart approach is still missing for accelerating ViT models. To close the above gap, we propose a first-of-its-kind algorithm-hardware codesigned framework, dubbed ViTALiTy, for boosting the inference efficiency of ViTs. Unlike sparsity-based Transformer accelerators for NLP, ViTALiTy unifies both low-rank and sparse components of the attention in ViTs. At the algorithm level, we approximate the dot-product softmax operation via first-order Taylor attention with row-mean centering as the low-rank component to linearize the cost of attention blocks and further boost the accuracy by incorporating a sparsity-based regularization. At the hardware level, we develop a dedicated accelerator to better leverage the resulting workload and pipeline from ViTALiTy’s linear Taylor attention which requires the execution of only the low-rank component, to further boost the hardware efficiency. Extensive experiments and ablation studies validate that ViTALiTy offers boosted end-to-end efficiency (e.g., $3\times$ faster and $3\times$ energy-efficient) under comparable accuracy, with respect to the state-of-the-art solution.
PDF 14 pages, 15 figures, Accepted to IEEE HPCA 2023
点此查看论文截图
Pure Transformer with Integrated Experts for Scene Text Recognition
Authors:Yew Lee Tan, Adams Wai-kin Kong, Jung-Jae Kim
Scene text recognition (STR) involves the task of reading text in cropped images of natural scenes. Conventional models in STR employ convolutional neural network (CNN) followed by recurrent neural network in an encoder-decoder framework. In recent times, the transformer architecture is being widely adopted in STR as it shows strong capability in capturing long-term dependency which appears to be prominent in scene text images. Many researchers utilized transformer as part of a hybrid CNN-transformer encoder, often followed by a transformer decoder. However, such methods only make use of the long-term dependency mid-way through the encoding process. Although the vision transformer (ViT) is able to capture such dependency at an early stage, its utilization remains largely unexploited in STR. This work proposes the use of a transformer-only model as a simple baseline which outperforms hybrid CNN-transformer models. Furthermore, two key areas for improvement were identified. Firstly, the first decoded character has the lowest prediction accuracy. Secondly, images of different original aspect ratios react differently to the patch resolutions while ViT only employ one fixed patch resolution. To explore these areas, Pure Transformer with Integrated Experts (PTIE) is proposed. PTIE is a transformer model that can process multiple patch resolutions and decode in both the original and reverse character orders. It is examined on 7 commonly used benchmarks and compared with over 20 state-of-the-art methods. The experimental results show that the proposed method outperforms them and obtains state-of-the-art results in most benchmarks.
PDF Accepted in ECCV2022
点此查看论文截图