Enabling DiffDock-PP on A100 GPU

DiffDock-PP provides new ways of analysing protein protein interactions, compared to original diffdock.

Unfortunately, the original DiffDock-PP build instructions do not work anymore as of Dec 2023.

This commit (10c6bb8b6e8279b5ec19e71e4b90e8963f10c53c) enables builds with more recent pytorch packages particularly pytorch geometric and CUDA 12.x, compared to the original diffdock-pp that used CUDA 11.6

conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidiaconda install pytorch-scatter -c pygconda install pytorch-sparse -c pygconda install pytorch-cluster -c pygconda install pytorch-spline-conv -c pygconda install pyg -c pygpip install numpy dill tqdm pyyaml pandas biopandas scikit-learn biopython e3nn wandb tensorboard tensorboardX matplotlibpip install numpy dill tqdm pyyaml pandas biopandas scikit-learn biopython e3nn wandb tensorboard tensorboardX matplotlib

08:03:08 Finished run test_large_model_dipstemp sampling, temp_psi, temp_sigma_data_tr, temp_sigma_data_rot: (2.439, 0.216, 0.593, 0.228)filtering_model_path: checkpoints/confidence_model_dips/fold_0/Total time spent: 2571.484815597534ligand_rmsd_summarized: {‘mean’: 28.133606, ‘median’: 17.168114, ‘std’: 27.295364, ‘lt1’: 0.0, ‘lt2’: 3.0, ‘lt5’: 37.0, ‘lt10’: 47.0}complex_rmsd_summarized: {‘mean’: 11.627805172380722, ‘median’: 7.572624206542969, ‘std’: 11.257617541135462, ‘lt1’: 7.0, ‘lt2’: 33.0, ‘lt5’: 48.0, ‘lt10’: 52.0}interface_rmsd_summarized: {‘mean’: 9.93406080286288, ‘median’: 6.356907844543457, ‘std’: 11.290295092623657, ‘lt1’: 8.0, ‘lt2’: 36.0, ‘lt5’: 48.0, ‘lt10’: 57.0}

08:03:12 Dumped data!! in storage/test_large_model_dips.pkl

The above commit has more specific details on the changes.