Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/deepwisdom/autodl
Automated Deep Learning without ANY human intervention. 1'st Solution for AutoDL challenge@NeurIPS.
https://github.com/deepwisdom/autodl
ai artificial-intelligence autodl autodl-challenge automated-machine-learning automl big-data data-science deeplearning feature-engineering full-automl lightgbm machine-learning model-selection multi-label nas python pytorch resnet tensorflow
Last synced: 22 days ago
JSON representation
Automated Deep Learning without ANY human intervention. 1'st Solution for AutoDL challenge@NeurIPS.
- Host: GitHub
- URL: https://github.com/deepwisdom/autodl
- Owner: DeepWisdom
- License: apache-2.0
- Created: 2020-04-02T14:59:04.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2022-09-23T22:40:53.000Z (about 2 years ago)
- Last Synced: 2024-10-15T10:05:12.216Z (28 days ago)
- Topics: ai, artificial-intelligence, autodl, autodl-challenge, automated-machine-learning, automl, big-data, data-science, deeplearning, feature-engineering, full-automl, lightgbm, machine-learning, model-selection, multi-label, nas, python, pytorch, resnet, tensorflow
- Language: Python
- Homepage: http://fuzhi.ai
- Size: 4.46 MB
- Stars: 1,135
- Watchers: 32
- Forks: 215
- Open Issues: 24
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
[English](./README_EN.md) | 简体中文
[![HitCount](http://hits.dwyl.com/DeepWisdom/AutoDL.svg)](http://hits.dwyl.com/DeepWisdom/AutoDL)
![GitHub All Releases](https://img.shields.io/github/downloads/DeepWisdom/AutoDL/total)
[![GitHub issues](https://img.shields.io/github/issues/DeepWisdom/AutoDL)](https://github.com/DeepWisdom/AutoDL/issues)
![GitHub closed issues](https://img.shields.io/github/issues-closed/DeepWisdom/AutoDL)
[![GitHub forks](https://img.shields.io/github/forks/DeepWisdom/AutoDL)](https://github.com/DeepWisdom/AutoDL/network)
[![GitHub stars](https://img.shields.io/github/stars/DeepWisdom/AutoDL)](https://github.com/DeepWisdom/AutoDL/stargazers)
![GitHub release (latest by date)](https://img.shields.io/github/v/release/deepwisdom/AutoDL)
[![GitHub license](https://img.shields.io/github/license/DeepWisdom/AutoDL)](https://github.com/DeepWisdom/AutoDL/blob/master/LICENSE)
![img](https://img.shields.io/badge/python-3.5-brightgreen)
[![img](https://img.shields.io/badge/chat-wechat-green)](https://github.com/DeepWisdom/AutoDL#%E5%8A%A0%E5%85%A5%E7%A4%BE%E5%8C%BA)![img](assets/autodl_logo_full.png)
[AutoDL Challenge@NeurIPS](https://autodl.chalearn.org/neurips2019) 冠军方案,竞赛细节参见 [AutoDL Competition](https://autodl.lri.fr/competitions/162)。
# 1. AutoDL是什么?
AutoDL聚焦于自动进行任意模态(图像、视频、语音、文本、表格数据)多标签分类的通用算法,可以用一套标准算法流解决现实世界的复杂分类问题,解决调数据、特征、模型、超参等烦恼,最短10秒就可以做出性能优异的分类器。本工程在**不同领域的24个离线数据集、15个线上数据集都获得了极为优异的成绩**。AutoDL拥有以下特性:
☕ **全自动**:全自动深度学习/机器学习框架,全流程无需人工干预。数据、特征、模型的所有细节都已调节至最佳,统一解决了资源受限、数据倾斜、小数据、特征工程、模型选型、网络结构优化、超参搜索等问题。**只需要准备数据,开始AutoDL,然后喝一杯咖啡**。
🌌 **通用性**:支持**任意**模态,包括图像、视频、音频、文本和结构化表格数据,支持**任意多标签分类问题**,包括二分类、多分类、多标签分类。它在**不同领域**都获得了极其优异的成绩,如行人识别、行人动作识别、人脸识别、声纹识别、音乐分类、口音分类、语言分类、情感分类、邮件分类、新闻分类、广告优化、推荐系统、搜索引擎、精准营销等等。
👍 **效果出色**:AutoDL竞赛获得压倒性优势的冠军方案,包含对传统机器学习模型和最新深度学习模型支持。模型库包括从LR/SVM/LGB/CGB/XGB到ResNet*/MC3/DNN/ThinResnet*/TextCNN/RCNN/GRU/BERT等优选出的冠军模型。
⚡ **极速/实时**:最快只需十秒即可获得极具竞争力的模型性能。结果实时刷新(秒级),无需等待即可获得模型实时效果反馈。
# 2. 目录
- [1. AutoDL是什么?](#1-autodl是什么)
- [2. 目录](#2-目录)
- [3. 效果](#3-效果)
- [4. AutoDL竞赛使用说明](#4-autodl竞赛使用说明)
- [4.1. 使用效果示例(横轴为对数时间轴,纵轴为AUC)](#41-使用效果示例横轴为对数时间轴纵轴为auc)
- [5. 安装](#5-安装)
- [5.1. pip 安装](#51-pip-安装)
- [6. 快速上手](#6-快速上手)
- [6.1. 快速上手之AutoDL本地效果测试](#61-快速上手之autodl本地效果测试)
- [6.2. 快速上手之图像分类](#62-快速上手之图像分类)
- [6.3. 快速上手之视频分类](#63-快速上手之视频分类)
- [6.4. 快速上手之音频分类](#64-快速上手之音频分类)
- [6.5. 快速上手之文本分类](#65-快速上手之文本分类)
- [6.6. 快速上手之表格分类](#66-快速上手之表格分类)
- [7. 可用数据集](#7-可用数据集)
- [7.1. (可选) 下载数据集](#71-可选-下载数据集)
- [7.2. 公共数据集信息](#72-公共数据集信息)
- [8. 贡献代码](#8-贡献代码)
- [9. 加入社区](#9-加入社区)
- [10. 开源协议](#10-开源协议)# 3. 效果
- **预赛榜单(DeepWisdom总分第一,平均排名1.2,在5个数据集中取得了4项第一)**
![img](assets/feedback-lb.png)- **决赛榜单(DeepWisdom总分第一,平均排名1.8,在10个数据集中取得了7项第一)**
![img](assets/final-lb-visual.png)# 4. AutoDL竞赛使用说明
1. 基础环境
```shell script
python>=3.5
CUDA 10
cuDNN 7.5
```2. clone仓库
```
cd
git clone https://github.com/DeepWisdom/AutoDL.git
```
3. 预训练模型准备
下载模型 [speech_model.h5](https://github.com/DeepWisdom/AutoDL/releases/download/opensource/thin_resnet34.h5) 放至 `AutoDL_sample_code_submission/at_speech/pretrained_models/` 目录。4. 可选:使用与竞赛同步的docker环境
- CPU
```
cd path/to/autodl/
docker run -it -v "$(pwd):/app/codalab" -p 8888:8888 evariste/autodl:cpu-latest
```
- GPU
```
nvidia-docker run -it -v "$(pwd):/app/codalab" -p 8888:8888 evariste/autodl:gpu-latest
```
5. 数据集准备:使用 `AutoDL_sample_data` 中样例数据集,或批量下载竞赛公开数据集。6. 进行本地测试
```
python run_local_test.py
```
本地测试完整使用。
```
python run_local_test.py -dataset_dir='AutoDL_sample_data/miniciao' -code_dir='AutoDL_sample_code_submission'
```
您可在 `AutoDL_scoring_output/` 目录中查看实时学习曲线反馈的HTML页面。细节可参考 [AutoDL Challenge official starting_kit](https://github.com/zhengying-liu/autodl_starting_kit_stable).
## 4.1. 使用效果示例(横轴为对数时间轴,纵轴为AUC)
![img](assets/AutoDL-performance-example.png)
可以看出,在五个不同模态的数据集下,AutoDL算法流都获得了极为出色的全时期效果,可以在极短的时间内达到极高的精度。
# 5. 安装
本仓库在 Python 3.6+, PyTorch 1.3.1 和 TensorFlow 1.15上测试.
你应该在[虚拟环境](https://docs.python.org/3/library/venv.html) 中安装autodl。
如果对虚拟环境不熟悉,请看 [用户指导](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/).用合适的Python版本创建虚拟环境,然后激活它。
## 5.1 windows10 安装过程
### 5.1.1 安装 cuda 10.0 和 cudnn v7.6.2.24
- [CUDA 10.0下载](https://developer.nvidia.com/cuda-10.0-download-archive?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal)
- [cuDNN下载](https://developer.nvidia.com/rdp/cudnn-archive)
- [百度云](https://pan.baidu.com/s/1BDP2gD7s-R0mcwXcFVe5Wg) 提取码:xb9x### 5.1.2 安装 Miniconda3-4.5.4-Windows-x86_64.exe
- [Miniconda3-4.5.4-Windows-x86_64.exe](https://repo.anaconda.com/miniconda/Miniconda3-4.5.4-Windows-x86_64.exe)
- [百度云](https://pan.baidu.com/s/1BDP2gD7s-R0mcwXcFVe5Wg) 提取码:xb9x### 5.1.3 安装 visualcppbuildtools_full.exe
- [visualcppbuildtools_full.exe](http://go.microsoft.com/fwlink/?LinkId=691126)
- [百度云](https://pan.baidu.com/s/1BDP2gD7s-R0mcwXcFVe5Wg) 提取码:xb9x### 5.1.4 创建 `start_env.bat` 文件
- 将其移动到安装的 `Miniconda3` 同级目录下
```bash
cmd.exe "/K" .\Miniconda3\Scripts\activate.bat .\Miniconda3
```### 5.1.5 双击 `start_env.bat` 安装 autodl-gpu
```bash
conda install pytorch==1.3.1
conda install torchvision -c pytorch
pip install autodl-gpu
```
## 5.2 Linux安装
```bash
pip install autodl-gpu
```# 6. 快速上手
## 6.1. 快速上手之AutoDL本地效果测试
指导参见 [快速上手之AutoDL本地效果测试](https://github.com/DeepWisdom/AutoDL/tree/pip/docs/run_local_test_tutorial_chn.md),样例代码参见 [examples/run_local_test.py](https://github.com/DeepWisdom/AutoDL/blob/pip/examples/run_local_test.py)## 6.2. 快速上手之图像分类
参见 [快速上手之图像分类](https://github.com/DeepWisdom/AutoDL/tree/pip/docs/image_classification_tutorial_chn.md),样例代码参见 [examples/run_image_classification_example.py](https://github.com/DeepWisdom/AutoDL/blob/pip/examples/run_image_classification_example.py)## 6.3. 快速上手之视频分类
指导参见 [快速上手之视频分类](https://github.com/DeepWisdom/AutoDL/tree/pip/docs/video_classification_tutorial_chn.md),样例代码参见[examples/run_video_classification_example.py](https://github.com/DeepWisdom/AutoDL/blob/pip/examples/run_video_classification_example.py)## 6.4. 快速上手之音频分类
指导参见 [快速上手之音频分类](https://github.com/DeepWisdom/AutoDL/tree/pip/docs/speech_classification_tutorial_chn.md),样例代码参见[examples/run_speech_classification_example.py](https://github.com/DeepWisdom/AutoDL/blob/pip/examples/run_speech_classification_example.py)## 6.5. 快速上手之文本分类
指导参见 [快速上手之文本分类](https://github.com/DeepWisdom/AutoDL/tree/pip/docs/text_classification_tutorial_chn.md),样例代码参见[examples/run_text_classification_example.py](https://github.com/DeepWisdom/AutoDL/blob/pip/examples/run_text_classification_example.py)。## 6.6. 快速上手之表格分类
指导参见 [快速上手之表格分类](https://github.com/DeepWisdom/AutoDL/tree/pip/docs/tabular_classification_tutorial_chn.md),样例代码参见[examples/run_tabular_classification_example.py](https://github.com/DeepWisdom/AutoDL/blob/pip/examples/run_tabular_classification_example.py).# 7. 可用数据集
## 7.1. (可选) 下载数据集
```bash
python download_public_datasets.py
```## 7.2. 公共数据集信息
| # | Name | Type | Domain | Size | Source | Data (w/o test labels) | Test labels |
| --- | -------- | ------- | -------- | ------ | ----------- | ---------------------- | ----------------- |
| 1 | Munster | Image | HWR | 18 MB | MNIST | munster.data | munster.solution |
| 2 | City | Image | Objects | 128 MB | Cifar-10 | city.data | city.solution |
| 3 | Chucky | Image | Objects | 128 MB | Cifar-100 | chucky.data | chucky.solution |
| 4 | Pedro | Image | People | 377 MB | PA-100K | pedro.data | pedro.solution |
| 5 | Decal | Image | Aerial | 73 MB | NWPU VHR-10 | decal.data | decal.solution |
| 6 | Hammer | Image | Medical | 111 MB | Ham10000 | hammer.data | hammer.solution |
| 7 | Kreatur | Video | Action | 469 MB | KTH | kreatur.data | kreatur.solution |
| 8 | Kreatur3 | Video | Action | 588 MB | KTH | kreatur3.data | kreatur3.solution |
| 9 | Kraut | Video | Action | 1.9 GB | KTH | kraut.data | kraut.solution |
| 10 | Katze | Video | Action | 1.9 GB | KTH | katze.data | katze.solution |
| 11 | data01 | Speech | Speaker | 1.8 GB | -- | data01.data | data01.solution |
| 12 | data02 | Speech | Emotion | 53 MB | -- | data02.data | data02.solution |
| 13 | data03 | Speech | Accent | 1.8 GB | -- | data03.data | data03.solution |
| 14 | data04 | Speech | Genre | 469 MB | -- | data04.data | data04.solution |
| 15 | data05 | Speech | Language | 208 MB | -- | data05.data | data05.solution |
| 16 | O1 | Text | Comments | 828 KB | -- | O1.data | O1.solution |
| 17 | O2 | Text | Emotion | 25 MB | -- | O2.data | O2.solution |
| 18 | O3 | Text | News | 88 MB | -- | O3.data | O3.solution |
| 19 | O4 | Text | Spam | 87 MB | -- | O4.data | O4.solution |
| 20 | O5 | Text | News | 14 MB | -- | O5.data | O5.solution |
| 21 | Adult | Tabular | Census | 2 MB | Adult | adult.data | adult.solution |
| 22 | Dilbert | Tabular | -- | 162 MB | -- | dilbert.data | dilbert.solution |
| 23 | Digits | Tabular | HWR | 137 MB | MNIST | digits.data | digits.solution |
| 24 | Madeline | Tabular | -- | 2.6 MB | -- | madeline.data | madeline.solution |# 8. 贡献代码
❤️ 请毫不犹豫参加贡献 [Open an issue](https://github.com/DeepWisdom/AutoDL/issues/new) 或提交 PRs。
# 9. 加入社区
# 10. 开源协议
[Apache License 2.0](https://github.com/DeepWisdom/AutoDL/blob/master/LICENSE)