Paddle pit - Rethinking Spatial Dimensions of Vision Transformers

Overview

基于Paddle实现PiT ——Rethinking Spatial Dimensions of Vision Transformers,arxiv

  • 官方原版代码(基于PyTorch)pit.

  • 本项目基于 PaddleViT 实现,在其基础上与原版代码实现了更进一步的对齐,并通过完整训练与测试完成对pit_ti模型的复现.

1. 简介

从CNN的成功设计原理出发,作者研究了空间尺寸转换的作用及其在基于Transformer的体系结构上的有效性。

具体来说,类似于CNN的降维原则(随着深度的增加,传统的CNN会增加通道尺寸并减小空间尺寸),作者用实验表明了这同样有利于Transformer的性能提升,并提出了基于池化的Vision Transformer,即PiT(模型示意图如下)。

drawing

PiT 模型示意图

2. 数据集和复现精度

数据集

原文使用的为ImageNet-1k 2012(ILSVRC2012),共1000类,训练集/测试集图片分布:1281167/50000,数据集大小为144GB。

本项目使用的为官方推荐的图片压缩过的更轻量的Light_ILSVRC2012,数据集大小为65GB。其在AI Studio上的地址为:Light_ILSVRC2012_part_0.tarLight_ILSVRC2012_part_1.tar

复现精度

Model 目标精度[email protected] 实现精度[email protected] Image Size batch_size Crop_pct epoch #Params
pit_ti 73.0 73.01 224 256*4GPUs 0.9 300
(+10 COOLDOWN)
4.8M

【注】上表中的实现精度在原版ILSVRC2012验证集上测试得到。 值得一提的是,本项目在Light_ILSVRC2012的验证集上的Validation [email protected]达到了73.17

本项目训练得到的最佳模型参数与训练日志log均存放于output文件夹下。

日志文件说明

本项目通过AI Studio的脚本任务运行,中途中断了4次,因此共有5个日志文件。为了方便检阅,本人手动将log命名为log_开始epoch-结束epoch.txt格式。具体来说:

  • output/log_1-76.txt:epoch1~epoch76。这一版代码定义每10个epoch保存一次模型权重,每2个epoch验证一次,同时若验证精度高于历史精度,则保存为Best_PiT.pdparams,因此在epoch76训练结束但还未验证的时候中断,下一次的训练只能从验证精度最高的epoch74继续训练。

  • output/log_75-142.txt:epoch75~epoch142。从这一版代码开始,新增了每次训练之后都保存一下模型参数为PiT-Latest.pdparams,这样无论哪个epoch训练中断都可以继续训练啦。

  • output/log_143-225.txt:epoch143~epoch225。

  • output/log_226-303.txt:epoch226~epoch303。

  • output/log_304-310.txt:epoch304~epoch310。

  • output/log_eval.txt:使用训练得到的最好模型(epoch308)在原版ILSVRC2012验证集上验证日志。

3. 准备环境

推荐环境配置:

本人环境配置:

  • 硬件:Tesla V100 * 4(由衷感谢百度飞桨平台提供高性能算力支持)

  • PaddlePaddle==2.2.1

  • Python==3.7

4. 快速开始

本项目现已通过脚本任务形式部署到AI Studio上,您可以选择fork下来直接运行sh run.sh,数据集处理等脚本均已部署好。链接:paddle_pit

或者您也可以git本repo在本地运行:

第一步:克隆本项目

git clone https://github.com/hatimwen/paddle_pit.git
cd paddle_pit

第二步:修改参数

请根据实际情况,修改scripts路径下的脚本内容(如:gpu,数据集路径data_path,batch_size等)。

第三步:验证模型

多卡请运行:

sh scripts/run_eval_multi.sh

单卡请运行:

sh scripts/run_eval.sh

第四步:训练模型

多卡请运行:

sh scripts/run_train_multi.sh

单卡请运行:

sh scripts/run_train.sh

第五步:验证预测

python predict.py \
-pretrained='output/Best_PiT' \
-img_path='images/ILSVRC2012_val_00004506.JPEG'

验证图片(类别:藏獒, id: 244)

输出结果为:

class_id: 244, prob: 9.12291145324707

对照ImageNet类别id(ImageNet数据集编号对应的类别内容),可知244为藏獒,预测结果正确。

5.代码结构

|-- paddle_pit
    |-- output              # 日志及模型文件
    |-- configs             # 参数
        |-- pit_ti.yaml
    |-- datasets
        |-- ImageNet1K      # 数据集路径
    |-- scripts             # 运行脚本
        |-- run_train.sh
        |-- run_train_multi.sh
        |-- run_eval.sh
        |-- run_eval_multi.sh
    |-- augment.py          # 数据增强
    |-- config.py           # 最底层配置文件
    |-- datasets.py         # dataset与dataloader
    |-- droppath.py         # droppath定义
    |-- losses.py           # loss定义
    |-- main_multi_gpu.py   # 多卡训练测试代码
    |-- main_single_gpu.py  # 单卡训练测试代码
    |-- mixup.py            # mixup定义
    |-- model_ema.py        # EMA定义
    |-- pit.py              # pit模型结构定义
    |-- random_erasing.py   # random_erasing定义
    |-- regnet.py           # 教师模型定义(本项目并未对此验证,仅作保留)
    |-- transforms.py       # RandomHorizontalFlip定义
    |-- utils.py            # CosineLRScheduler及AverageMeter定义
    |-- README.md
    |-- requirements.txt

6. 参考及引用

@InProceedings{Yuan_2021_ICCV,
    author    = {Yuan, Li and Chen, Yunpeng and Wang, Tao and Yu, Weihao and Shi, Yujun and Jiang, Zi-Hang and Tay, Francis E.H. and Feng, Jiashi and Yan, Shuicheng},
    title     = {Tokens-to-Token ViT: Training Vision Transformers From Scratch on ImageNet},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {558-567}
}

最后,非常感谢百度举办的飞桨论文复现挑战赛(第五期)让本人对Paddle理解更加深刻。 同时也非常感谢朱欤老师团队用Paddle实现的PaddleViT,本项目大部分代码都是从中copy来的,而仅仅实现了其与原版代码训练步骤的进一步对齐与完整的训练过程,但本人也同样受益匪浅! ♥️

Contact

Owner
Hongtao Wen
Hongtao Wen
A Context-aware Visual Attention-based training pipeline for Object Detection from a Webpage screenshot!

CoVA: Context-aware Visual Attention for Webpage Information Extraction Abstract Webpage information extraction (WIE) is an important step to create k

Keval Morabia 41 Jan 01, 2023
Open source Python implementation of the HDR+ photography pipeline

hdrplus-python Open source Python implementation of the HDR+ photography pipeline, originally developped by Google and presented in a 2016 article. Th

77 Jan 05, 2023
iris - Open Source Photos Platform Powered by PyTorch

Open Source Photos Platform Powered by PyTorch. Submission for PyTorch Annual Hackathon 2021.

Omkar Prabhu 137 Sep 10, 2022
Train a state-of-the-art yolov3 object detector from scratch!

TrainYourOwnYOLO: Building a Custom Object Detector from Scratch This repo let's you train a custom image detector using the state-of-the-art YOLOv3 c

AntonMu 616 Jan 08, 2023
The official implementation of the Hybrid Self-Attention NEAT algorithm

PUREPLES - Pure Python Library for ES-HyperNEAT About This is a library of evolutionary algorithms with a focus on neuroevolution, implemented in pure

Adrian Westh 91 Dec 12, 2022
HyDiff: Hybrid Differential Software Analysis

HyDiff: Hybrid Differential Software Analysis This repository provides the tool and the evaluation subjects for the paper HyDiff: Hybrid Differential

Yannic Noller 22 Oct 20, 2022
Deep Markov Factor Analysis (NeurIPS2021)

Deep Markov Factor Analysis (DMFA) Codes and experiments for deep Markov factor analysis (DMFA) model accepted for publication at NeurIPS2021: A. Farn

Sarah Ostadabbas 2 Dec 16, 2022
Augmented CLIP - Training simple models to predict CLIP image embeddings from text embeddings, and vice versa.

Train aug_clip against laion400m-embeddings found here: https://laion.ai/laion-400-open-dataset/ - note that this used the base ViT-B/32 CLIP model. S

Peter Baylies 55 Sep 13, 2022
Implementation for ACProp ( Momentum centering and asynchronous update for adaptive gradient methdos, NeurIPS 2021)

This repository contains code to reproduce results for submission NeurIPS 2021, "Momentum Centering and Asynchronous Update for Adaptive Gradient Meth

Juntang Zhuang 15 Jun 11, 2022
Based on the given clinical dataset, Predict whether the patient having Heart Disease or Not having Heart Disease

Heart_Disease_Classification Based on the given clinical dataset, Predict whether the patient having Heart Disease or Not having Heart Disease Dataset

Ashish 1 Jan 30, 2022
A PyTorch implementation of SIN: Superpixel Interpolation Network

SIN: Superpixel Interpolation Network This is is a PyTorch implementation of the superpixel segmentation network introduced in our PRICAI-2021 paper:

6 Sep 28, 2022
High-quality implementations of standard and SOTA methods on a variety of tasks.

Uncertainty Baselines The goal of Uncertainty Baselines is to provide a template for researchers to build on. The baselines can be a starting point fo

Google 1.1k Dec 30, 2022
This repository contains code accompanying the paper "An End-to-End Chinese Text Normalization Model based on Rule-Guided Flat-Lattice Transformer"

FlatTN This repository contains code accompanying the paper "An End-to-End Chinese Text Normalization Model based on Rule-Guided Flat-Lattice Transfor

THUHCSI 74 Nov 28, 2022
A package related to building quasi-fibration symmetries

qf A package related to building quasi-fibration symmetries. If you'd like to learn more about how it works, see the brief explanation and References

Paolo Boldi 1 Dec 01, 2021
Reinforcement learning models in ViZDoom environment

DoomNet DoomNet is a ViZDoom agent trained by reinforcement learning. The agent is a neural network that outputs a probability of actions given only p

Andrey Kolishchak 126 Dec 09, 2022
Reviving Iterative Training with Mask Guidance for Interactive Segmentation

This repository provides the source code for training and testing state-of-the-art click-based interactive segmentation models with the official PyTorch implementation

Visual Understanding Lab @ Samsung AI Center Moscow 406 Jan 01, 2023
Detector for Log4Shell exploitation attempts

log4shell-detector Detector for Log4Shell exploitation attempts Idea The problem with the log4j CVE-2021-44228 exploitation is that the string can be

Florian Roth 729 Dec 25, 2022
RoIAlign & crop_and_resize for PyTorch

RoIAlign for PyTorch This is a PyTorch version of RoIAlign. This implementation is based on crop_and_resize and supports both forward and backward on

Long Chen 530 Jan 07, 2023
A Research-oriented Federated Learning Library and Benchmark Platform for Graph Neural Networks. Accepted to ICLR'2021 - DPML and MLSys'21 - GNNSys workshops.

FedGraphNN: A Federated Learning System and Benchmark for Graph Neural Networks A Research-oriented Federated Learning Library and Benchmark Platform

FedML-AI 175 Dec 01, 2022
PyTorch implementation of Octave Convolution with pre-trained Oct-ResNet and Oct-MobileNet models

octconv.pytorch PyTorch implementation of Octave Convolution in Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octa

Duo Li 273 Dec 18, 2022