How to train a parallel model? #65
-
|
I want to train a model to do scalable MD simulation, how to do this? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
You don't need another training steps to run parallel, scalable MD simulations.
Check 'for parallel model' section. Note that currently parallel MD run can not compute virial pressure. Plus, cuda-aware openMPI is essential for optimal performance. You should choose number of the mpi process equals the number of GPUs you want to use. |
Beta Was this translation helpful? Give feedback.
You don't need another training steps to run parallel, scalable MD simulations.
Deploy your model via
sevenn_get_model -p {path to checkpoint}Run LAMMPS with
mpirunafter writing appropriate input script. You can find the guideline from here: https://github.com/MDIL-SNU/SevenNetCheck 'for parallel model' section.
Note that currently parallel MD run can not compute virial pressure. Plus, cuda-aware openMPI is essential for optimal performance. You should choose number of the mpi process equals the number of GPUs you want to use.