# GEGFM-Paddle

**Repository Path**: owhileo/gegfm-paddle

## Basic Information

- **Project Name**: GEGFM-Paddle
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-09-12
- **Last Updated**: 2025-09-12

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Code description

This is the c++ and Paddle implementation of our new graph encryption on private graph model inference.

## Requirements

- SEAL version 4.1
- OpenSSL

## Running the Code

```bash
cd cpp

g++ main.cpp graph_load.cpp graph_enc.cpp graph_he.cpp -o gegcn -I~/helibs/include/SEAL-4.1 -L~/helibs/lib -l:libseal-4.1.a -lssl -lcrypto -O2 

./gegcn 
```

Replace `~/helibs` with the path where `SEAL` library is installed.

```bash
cd py

python ./ProG/pre_train.py --pretrain_task Edgepred_GPPT --dataset_name PubMed --gnn_type GIN --hid_dim 128 --batch_size 256 --seed 42 --device 0 --patience 3 --epochs 75 --num_layer 2

python ./ProG/downstream_task.py --pre_train_model_path './Experiment/pre_trained_model/PubMed/Edgepred_GPPT.GIN.128hidden_dim.pth' --downstream_task NodeTask --dataset_name 'PubMed' --gnn_type 'GIN' --prompt_type 'GPF-plus'  --hid_dim 128 --lr 0.0005 --decay 2e-6 --seed 42 --device 0 --hop 8 --task_num 1 --batch_size 256 --smallest_size 1 --shot_num 1000 --num_layer 2 --pnum 50 --epochs 300

python ./ProG/neighbor_sampling_task.py --gnn_type GIN --pre_train_model_path './Experiment/model/PubMed/GIN/1000-shot_GPF-plus.pth' --dataset_name 'PubMed' --seed 42 --batch_size 256 --epsilon_min 0.3 --hop 8 --smallest_size 1 --use_geometric 1 --epsilon_max 3
```

ATTN: This package is free for academic usage. You can run it at your own risk. For other purposes, please contact yuanmj@lamda.nju.edu.cn.