How to Perform Parallel Inference

[itsuwari]

What would you like to report?

Hello,

FAIR-CHEM models are quite GPU-intensive. Is there an established method or best practice for performing inference on multiple GPUs to improve efficiency? Some examples in the document would be great.

Thank you!

[lbluque]

Hi @itsuwari,

Can you give use more details on what you are trying to do?

Do you want to perform parallel inference for a single system (ie a large supercell/many atoms)? or on many small systems (~100 atoms or less)?

Are you using the OCPCalculator or using the CLI?

[itsuwari]

Hello @lbluque,
I want to perform parallel inference for a single large system with approx. 1k atoms, with OCPCalculator.
An example for parallel inference would be great in the document!
Thanks.