Josephus Problem GNN Solver

This project uses a Graph Neural Network (GNN) to solve the Josephus problem. It includes training, validation, and inference capabilities.

Files

1. model.py: Defines the GNN model structure.
2. dataset.py: Handles data loading and preprocessing.
3. utils.py: Contains utility functions for training and evaluation.
4. train.py: Main script for training the model.
5. inference.py: Script for making predictions using a trained model.
6. requirements.txt: Lists all required Python packages.
7. josephus_train.csv: Training dataset.
8. josephus_test.csv: Testing dataset.

Setup

1. Ensure you have Python 3.7+ installed.

2. Install the required packages:

   pip install -r requirements.txt
   

Training

To train the model:

python train.py

This script will train the model using the data in josephus_train.csv and validate it using josephus_test.csv. The trained model will be saved as josephus_model.pth.

Inference

To make predictions using a trained model:

python inference.py

This script will load the trained model and allow you to input values for N and M to predict the survivor's position.

Distributed Training

The train.py script supports distributed training. To run on multiple GPUs:

torchrun --nproc_per_node=NUM_GPUS train.py

Replace NUM_GPUS with the number of GPUs you want to use.

Data

The josephus_train.csv and josephus_test.csv files contain the training and testing data respectively. Each row represents a Josephus problem instance with columns for N (number of people), M (skip count), and Result (survivor's position).

Note

Ensure you have sufficient GPU memory for training, especially for large datasets. Adjust batch sizes in train.py if needed.