BTTackler

A Diagnosis-based Framework for Efficient Deep Learning Hyperparameter Optimization.

Install

Based on Python 3.10

git clone git@github.com:Dmagine/BTTackler.git
cd BTTackler
pip install -r requirements.txt
pip install -U .

Run Cases

cd cases
bash example_run_case.sh

or just,

cd cases
# random
python run_case.py cifar10cnn random
# bttackler-random
python run_case.py cifar10cnn random_bttackler
# all for cifar10cnn
python run_case.py cifar10cnn all

TODO

1. Higher code readability and configurable quality indicators.
2. Code formatting and complete experiment cases.
3. Cases for calibrating quality indicators.

Citation

If you find BTTackler useful, please cite our paper.

@inproceedings{pei2024bttackler,
  title={BTTackler: A Diagnosis-based Framework for Efficient Deep Learning Hyperparameter Optimization},
  author={Zhongyi Pei and Zhiyao Cen and Yipeng Huang and Chen Wang and Lin Liu and Philip Yu and Mingsheng Long and Jianmin Wang},
  booktitle={In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  year={2024},
}

Contact

If you have any questions or suggestions, feel free to contact:

• Zhongyi Pei (peizhyi@tsinghua.edu.cn)
• Zhiyao Cen (cenzy23@mails.tsinghua.edu.cn)

Or describe it in Issues.