The Domino Open MPI base image contains common dependencies like Pandas and NumPy. You can modify your environments to add additional packages or a specific package version.
You can package and manage dependencies for your MPI-enabled compute environments. This creates flexibility in managing dependencies for individual projects or workloads without shared cluster worker complexities.
To add a new cluster worker, add the appropriate statements in the Docker Instructions section of both the MPI cluster and Open MPI execution environments. For example, to add a particular version of
mpi4pypackage include the following:
### Optionally specify version if desired RUN pip install PyTorch==1.9.0 RUN pip install mpi4py