AutoDL Projects
1.0.0
自动化深度学习项目(Autodl-Projects)是每个人的开源,轻巧但有用的项目。该项目实施了几种神经体系结构搜索(NAS)和高参数优化(HPO)算法。 中文介绍见readme_cn.md
谁应该考虑使用自动项目
我们为什么要使用autodl-projects
目前,该项目提供以下算法和脚本来运行它们。请参阅“描述”列中提供的链接中的详细信息。
| 类型 | abbrv | 算法 | 描述 |
|---|---|---|---|
| nas | tas | 通过可转换体系结构搜索进行网络修剪 | Neurips-2019-Tas.md |
| 飞镖 | 飞镖:可区分架构搜索 | ICLR-2019-darts.md | |
| gdas | 在四个GPU小时内搜索强大的神经建筑 | CVPR-2019-GDAS.MD | |
| 设定 | 通过自我评估的模板网络搜索单发神经架构搜索 | ICCV-2019-setn.md | |
| NAS Bench-2011 | NAS Bench-2011:扩展可再现神经体系结构搜索的范围 | NAS Bench-2011.md | |
| 纳斯板凳 | NATS板凳:为建筑拓扑和大小的基准测试NAS算法 | Nats-Bench.md | |
| ... | enas / rea / readforce / bohb | 请检查原始论文 | NAS-Bench-201.MD Nats-Bench.md |
| HPO | HPO-CG | 高参数优化,近似梯度 | 即将推出 |
| 基本的 | 重新连接 | 基于深度学习的图像分类 | 基线 |
首先,请使用pip install .安装xautodl库。
请安装Python>=3.6和PyTorch>=1.5.0 。 (您可以使用较低版本的Python和Pytorch,但可能有错误)。一些可视化代码可能需要opencv 。
Cifar和Imagenet应下载并提取到$TORCH_HOME中。一些方法使用知识蒸馏(KD),需要预训练的模型。请从Google Drive(或自己训练)下载这些模型,然后将其保存到.latent-data中。
请使用
git clone --recurse-submodules https://github.com/D-X-Y/AutoDL-Projects.git XAutoDL
使用子模型下载此回购。
如果您发现该项目有助于您的研究,请考虑引用相关论文:
@inproceedings{dong2021autohas,
title = {{AutoHAS}: Efficient Hyperparameter and Architecture Search},
author = {Dong, Xuanyi and Tan, Mingxing and Yu, Adams Wei and Peng, Daiyi and Gabrys, Bogdan and Le, Quoc V},
booktitle = {2nd Workshop on Neural Architecture Search at International Conference on Learning Representations (ICLR)},
year = {2021}
}
@article{dong2021nats,
title = {{NATS-Bench}: Benchmarking NAS Algorithms for Architecture Topology and Size},
author = {Dong, Xuanyi and Liu, Lu and Musial, Katarzyna and Gabrys, Bogdan},
doi = {10.1109/TPAMI.2021.3054824},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)},
year = {2021},
note = {mbox{doi}:url{10.1109/TPAMI.2021.3054824}}
}
@inproceedings{dong2020nasbench201,
title = {{NAS-Bench-201}: Extending the Scope of Reproducible Neural Architecture Search},
author = {Dong, Xuanyi and Yang, Yi},
booktitle = {International Conference on Learning Representations (ICLR)},
url = {https://openreview.net/forum?id=HJxyZkBKDr},
year = {2020}
}
@inproceedings{dong2019tas,
title = {Network Pruning via Transformable Architecture Search},
author = {Dong, Xuanyi and Yang, Yi},
booktitle = {Neural Information Processing Systems (NeurIPS)},
pages = {760--771},
year = {2019}
}
@inproceedings{dong2019one,
title = {One-Shot Neural Architecture Search via Self-Evaluated Template Network},
author = {Dong, Xuanyi and Yang, Yi},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV)},
pages = {3681--3690},
year = {2019}
}
@inproceedings{dong2019search,
title = {Searching for A Robust Neural Architecture in Four GPU Hours},
author = {Dong, Xuanyi and Yang, Yi},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {1761--1770},
year = {2019}
}
如果您想为此仓库做出贡献,请参阅parduting.md。此外,请遵循导电守则。
我们将black用于Python代码格式。请使用black . -l 88 。
整个代码库均在MIT许可下。