IA-GPL

View on GitHub

This repository contains the pytorch code for 'Instance-Aware Graph Prompt Learning' which is accepted in TMLR. We insert instance-aware prompts to improve performance and efficiency in downstream graph-related tasks.

Model Architecture

model figure

Setup

conda create -n IAGPL python=3.11
conda activate IAGPL
conda install -r requirements.txt

Datasets

We have provided 6 relatively smaller molecule datasets under dataset/ folder. Please download HIV and MUV datasets from repo and put them under the same folder.

Usage

python prompt_tuning.py [--device DEVICE] [--epochs EPOCHS] ...

For a complete list of hyperparameters, please check the arguments section in the prompt_tuning.py file.

Reproduction

We have provided scripts with hyper-parameter settings to reproduce the experimental results presented in our paper. You can simply run: bash run_final.sh

This codebase is based on GPF. The pre-trained GNNs and datasets are from repo. We thank these authors for their great works.