Project 3 - Dicemen

NOTE: Only model files are available because of copyright on original codebase.

Installation

1. Create the environment

   conda env create -f environment.yml

2. Activate the environment

   conda activate mp_project3

3. Run the following commands to install the remaining dependencies about pytorch-geometric in to the activated environment

   pip install torch-scatter -f https://pytorch-geometric.com/whl/torch-1.7.0+cu101.html
   pip install torch-sparse -f https://pytorch-geometric.com/whl/torch-1.7.0+cu101.html
   pip install torch-cluster -f https://pytorch-geometric.com/whl/torch-1.7.0+cu101.html
   pip install torch-spline-conv -f https://pytorch-geometric.com/whl/torch-1.7.0+cu101.html
   pip install torch-geometric

   Note: If pip commands give error, it is necessary to follow these steps below:

   wget https://pytorch-geometric.com/whl/torch-1.7.0+cu101/torch_cluster-1.5.9-cp38-cp38-linux_x86_64.whl
   wget https://pytorch-geometric.com/whl/torch-1.7.0+cu101/torch_scatter-2.0.6-cp38-cp38-linux_x86_64.whl
   wget https://pytorch-geometric.com/whl/torch-1.7.0+cu101/torch_sparse-0.6.9-cp38-cp38-linux_x86_64.whl
   wget https://pytorch-geometric.com/whl/torch-1.7.0+cu101/torch_spline_conv-1.2.1-cp38-cp38-linux_x86_64.whl
   pip install torch-geometric
   pip install wheel
   pip install torch_cluster-1.5.9-cp38-cp38-linux_x86_64.whl
   pip install torch_scatter-2.0.6-cp38-cp38-linux_x86_64.whl
   pip install torch_sparse-0.6.9-cp38-cp38-linux_x86_64.whl
   pip install torch_spline_conv-1.2.1-cp38-cp38-linux_x86_64.whl

4. Submit the training task to GPU with the following command (indicated time necessary to reproduce results)

   cd codebase/
   bsub -n 4 -W 24:00 -o sample_test -R "rusage[mem=4096, ngpus_excl_p=1]" -R "select[gpu_mtotal0>=10240]" python train.py ../configs/convgatadv.yaml

5. Submit the prediction task to GPU with the following command

   bsub -n 4 -W 2:00 -o sample_test -R "rusage[mem=4096, ngpus_excl_p=1]" -R "select[gpu_mtotal0>=10240]" python test.py ../configs/convgatadv.yaml --gen_model_file gen_model_200000.pt --disc_model_file disc_model_200000.pt

Results will be saved under a directory with the same name as the model (NOT in the top directory as the vanilla code did).