Installs

pip install torch_geometric pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.0.0+cu118.html pip install pytorch-lightning cd FAST-GNN-SAT/PyMiniSolvers/ !make cd ..

Generate data

To generate random dataset run:

1_generate_problems.sh 

This will create random train, validation and test data.

Train or finetune

Model is trained with:

2_train.sh

This will train the model and save he checkpoint. It is the way result in Fig 2 in the paper was obtained. The script runs the experiment with curriculum. To run the experiment without curriculum, uncomment the second line and comment out the first.

Compute cluster centers and test

After model is trained run:

3_copmute_cluster_centers.sh

Computes cluster centers for true and false

and

4_test.sh

Will use the trained model to evaluate the result (only for SR(40) for other datasets uncomment the other lines in the script).