Skip to content

Official code for our CVPR'22 paper “Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-learning”

License

Notifications You must be signed in to change notification settings

transmuteAI/MetaDOCK

Repository files navigation

MetaDOCK

This is the official repository to the CVPR 2022 paper "Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-learning" This repo is based on the training code in iMAML.

Getting Started

You will need Python 3.8 and the packages specified in requirements.txt. We recommend setting up a virtual environment with pip and installing the packages there.

Install packages with:

$ pip install -r requirements.txt

Configure and Run

All configurations concerning data, model, training, etc. can be called using commandline arguments.

Training

The implicit_maml script offers many options to train implicit-MAML on 4-conv model family and cifar-fs/miniimagenet dataset.

Here is a sample script to train on cifar-fs dataset, 4-conv model.

python implicit_maml.py 

Training

The implicit_maml_pruner script offers many options to prune the pre-trained model at different budgets.

Here is a sample script to prune on cifar-fs dataset, 4-conv model.

python implicit_maml_pruner.py

Citation

Please cite our paper in your publications if it helps your research.

@inproceedings{chavan2022metadock,
  title={Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-learning},
  author={Chavan, Arnav and Tiwari, Rishabh and Bamba, Udbhav and Gupta, Deepak},
  journal={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2022}
}

License

This project is licensed under the MIT License.

About

Official code for our CVPR'22 paper “Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-learning”

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages