idyllic990920 / FFL-pytorch

Feature Fusion for Online Mutual Knowledge Distillation Code

Feature Fusion for Online Mutual Knowledge Distillation

This repository is the official implementation of Feature Fusion for Online Mutual Knowledge Distillation (FFL). The source code is for reproducing the results of Table 1 of the original paper.

Requirements

To install requirements using environment.yml refer to the documentation.

Training

train_FFL.py is the code for training FFL with Mutual Knowledge Distillation (MKD). To train the model(s) in the paper, run this command:

#The results from the original paper can be reproducd by running : 
python train_FFL.py  --lr 0.1 --cu_num 0 --depth 32

Citation

Please refer to the following citation if this repository is useful for your research.

Bibtex:

@article{kim2019feature,
  title={Feature fusion for online mutual knowledge distillation},
  author={Kim, Jangho and Hyun, Minsung and Chung, Inseop and Kwak, Nojun},
  journal={arXiv preprint arXiv:1904.09058},
  year={2019}
}