klightz / Firefly

Offical Repo for Firefly Neural Architecture Descent: a General Approach for Growing Neural Networks. Accepted by Neurips 2020.

Firefly Neural Architecture Descent: a General Approach for Growing Neural Networks

This repository is the official implementation of Firefly Neural Architecture Descent: a General Approach for Growing Neural Networks

This work is received by NeurIPS 2020.

Requirements

To run the code, please download the pytorch >= 1.0 with torchvision

Training

To train the model(s) in the paper, run this command:

python main.py --method fireflyn --model vgg19

You can also try different growing method [exact/fast/fireflyn/random] which represent original splitting, fast splitting, firefly splitting, NASH described in the paper.

TODO

Detail tutorial and more experiments in the paper are under construction, will release in the future.