Skip to content

This repo contains project on graph KD done under Prof. Sharad Sinha

Notifications You must be signed in to change notification settings

k-styles/MS_CPF_KD

Repository files navigation

MS_CPF_KD

This project work combines the work of CPF student and MultiScale KD in GNNs papers. The references have been provided in the presentation of this project.

SETUP

Setup conda environment with enviornment.yml file

conda env create -file environment.yml

Training scripts:

python train_teacher.py --teacher GAT  --gpu 0 --dataset cora
python train_student.py --teacher GAT --student PLP --ptype ind --gpu 0 --dataset cora

The teachers will be trained in models/ directory, and afterwards you can train your student from the teachers that are there in the models/ directory.

NOTE: If you get an error, you've probably forgot to delete the models/ directory :)

This code should only be used for academic purposes.

About

This repo contains project on graph KD done under Prof. Sharad Sinha

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published