Adapting LLaMA Decoder to Vision Transformer

• URL: http://arxiv.org/abs/2404.06773v4
• Date: Mon, 27 May 2024 10:13:26 GMT
• Title: Adapting LLaMA Decoder to Vision Transformer
• Authors: Jiahao Wang, Wenqi Shao, Mengzhao Chen, Chengyue Wu, Yong Liu, Taiqiang Wu, Kaipeng Zhang, Songyang Zhang, Kai Chen, Ping Luo,
• Abstract summary: This work examines whether decoder-only Transformers such as LLaMA can be adapted to the computer vision field. We first "LLaMAfy" a standard ViT step-by-step to align with LLaMA's architecture, and find that directly applying a causal mask to the self-attention brings an attention collapse issue. We develop a soft mask strategy that gradually introduces a causal mask to the self-attention at the onset of training to facilitate the optimization behavior.
• Score: 65.47663195233802
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: This work examines whether decoder-only Transformers such as LLaMA, which were originally designed for large language models (LLMs), can be adapted to the computer vision field. We first "LLaMAfy" a standard ViT step-by-step to align with LLaMA's architecture, and find that directly applying a causal mask to the self-attention brings an attention collapse issue, resulting in the failure to the network training. We suggest to reposition the class token behind the image tokens with a post-sequence class token technique to overcome this challenge, enabling causal self-attention to efficiently capture the entire image's information. Additionally, we develop a soft mask strategy that gradually introduces a causal mask to the self-attention at the onset of training to facilitate the optimization behavior. The tailored model, dubbed as image LLaMA (iLLaMA), is akin to LLaMA in architecture and enables direct supervised learning. Its causal self-attention boosts computational efficiency and learns complex representation by elevating attention map ranks. iLLaMA rivals the performance with its encoder-only counterparts, achieving 75.1% ImageNet top-1 accuracy with only 5.7M parameters. Scaling the model to $\sim$310M and pre-training on ImageNet-21K further enhances the accuracy to 86.0%. Extensive experiments demonstrate iLLaMA's reliable properties: shape-texture bias, calibration, quantization compatibility, ADE20K segmentation and CIFAR transfer learning. We hope our study can kindle fresh views to visual architectures in the wave of LLMs and inspire the development of unified multimodal models. Pre-trained models and codes are available https://github.com/techmonsterwang/iLLaMA.

Related papers

• Bringing Masked Autoencoders Explicit Contrastive Properties for Point Cloud Self-Supervised Learning [116.75939193785143]
  Contrastive learning (CL) for Vision Transformers (ViTs) in image domains has achieved performance comparable to CL for traditional convolutional backbones. In 3D point cloud pretraining with ViTs, masked autoencoder (MAE) modeling remains dominant.
  arXiv  Detail & Related papers  (2024-07-08T12:28:56Z)
• Improve Supervised Representation Learning with Masked Image Modeling [30.30649867772395]
  We propose a simple yet effective setup that can easily integrate masked image modeling into existing supervised training paradigms. We show with minimal change in architecture and no overhead in inference that this setup is able to improve the quality of the learned representations.
  arXiv  Detail & Related papers  (2023-12-01T22:03:25Z)
• LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention [52.6718081345361]
  LLaMA-Adapter is a method to efficiently fine-tune LLaMA into an instruction-following model. It introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs.
  arXiv  Detail & Related papers  (2023-03-28T17:59:12Z)
• DeepMIM: Deep Supervision for Masked Image Modeling [46.01916629713594]
  Deep supervision was widely used in image classification in the early deep learning era. With the emergence of normalization techniques and residual connection, deep supervision in image classification was gradually phased out. We revisit deep supervision for masked image modeling (MIM) that pre-trains a Vision Transformer (ViT) via a mask-and-predict scheme.
  arXiv  Detail & Related papers  (2023-03-15T17:59:55Z)
• SdAE: Self-distillated Masked Autoencoder [95.3684955370897]
  Self-distillated masked AutoEncoder network SdAE is proposed in this paper. With only 300 epochs pre-training, a vanilla ViT-Base model achieves an 84.1% fine-tuning accuracy on ImageNet-1k classification.
  arXiv  Detail & Related papers  (2022-07-31T15:07:25Z)
• Adversarial Masking for Self-Supervised Learning [81.25999058340997]
  Masked image model (MIM) framework for self-supervised learning, ADIOS, is proposed. It simultaneously learns a masking function and an image encoder using an adversarial objective. It consistently improves on state-of-the-art self-supervised learning (SSL) methods on a variety of tasks and datasets.
  arXiv  Detail & Related papers  (2022-01-31T10:23:23Z)
• Masked Autoencoders Are Scalable Vision Learners [60.97703494764904]
  Masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. Coupling these two designs enables us to train large models efficiently and effectively.
  arXiv  Detail & Related papers  (2021-11-11T18:46:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.