Geometric Embedding Representations from Transformers. Implementation of GERT as described in Euclidean Transformers for Macromolecular Structures: Lessons Learned by DD Liu, L Melo, A Costa, M Vögele, RJL Townshend, and RO Dror.
Navigate to sbatches/install.sbatch and point the variable CONDA_BIN to your conda binary. You can run which conda to identify the location of your conda binary.
Then, navigate to the gert/src/atom3d_combined directory and run the following command to install:
sbatch ./sbatches/install.sbatch
This will allocate a GPU node and submit a batch script to SLURM. The batch script will handle:
- loading the appropriate CUDA/GCC modules,
- creating the conda environment called "gert" (if you want to name your conda environment to something else, change name declared in the environment.yml),
- installing the required libraries.
When the installation is done, activate the conda environment with conda activate <env_name>. You may also need to re-load the correct gcc and cuda versions.
We then need to separately install pytorch-geometric (described at https://github.com/pyg-team/pytorch_geometric). Use the commands
pip install torch-cluster==1.5.9 -f https://pytorch-geometric.com/whl/${TORCH_VERSION}+${CUDA}.html
pip install torch-geometric==1.7.0 -f https://pytorch-geometric.com/whl/${TORCH_VERSION}+${CUDA}.html
pip install torch-scatter==2.0.6 -f https://pytorch-geometric.com/whl/${TORCH_VERSION}+${CUDA}.html
pip install torch-sparse==0.6.9 -f https://pytorch-geometric.com/whl/${TORCH_VERSION}+${CUDA}.html
pip install torch-spline-conv==1.2.1 -f https://pytorch-geometric.com/whl/${TORCH_VERSION}+${CUDA}.html
For example, if you have torch 1.7.1 and cuda 11.0 (these version should work after the install script), use
pip install torch-cluster==1.5.9 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-geometric==1.7.0 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-scatter==2.0.6 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-sparse==0.6.9 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-spline-conv==1.2.1 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
etc.
Now, test that your installation works by running python3 run.py. There should be no error messages.
sbatch sbatches/1gpu.sbatch python command.py arg1 arg2 ...
Make a file like this:
vim ~/.gert
Populate that file with the following:
source <path to conda.sh>
conda deactivate
conda deactivate
module load cuda/10.1.168
conda activate <env_name>