Notes & FAQ¶
A grab-bag of tips collected from user issues and questions.
GPU / DGL¶
- Install a DGL build that matches your CUDA runtime, e.g.
pip install dgl-cu111for CUDA 11.1. Mismatched builds are the most common install failure. - If you see errors involving
libcudart.soorlibtorch_cuda.so, reinstall DGL from the wheel index that matches your PyTorch CUDA version (see Installation).
Structure file parsing¶
- Simple
.cifand.pdbfiles are handled byjarvis-toolsdirectly. - For more complex CIFs, install
cif2cell==2.0.0a3. - For complex PDBs, install
pytrajviaconda install -c ambermd pytraj.
Training hyperparameters¶
- The example
config_example.jsonships withbatch_size: 2so the test suite runs fast. Usebatch_size: 32or64for real trainings — otherwise training will be very slow and under-performing. pandas >= 1.2.3is required.- Starting March 2024,
pytorch-igniteis no longer a dependency (removed for conda-forge build compatibility).
CLIs are importable scripts¶
train_alignn.py, pretrained.py, and run_alignn_ff.py are installed as executables
in your environment's bin/ directory. You do not need the absolute path — just run
them.
Known dataset issues¶
- QM9 results: see issue #54 for details on a data-split discrepancy that affects reproducibility.
Getting help¶
- File a GitHub issue: https://github.com/atomgptlab/alignn/issues
- Email:
drkamal@jhu.edu