Identify and run exemplar model on Delta

[WillEngler]

To make sure we know how to use Delta and have a model in mind for our end-to-end Garden HPC demo, we need to identify a model and make sure we can run it manually, without Garden. Eliu mentioned gravitational wave detection and a diffusion model for ligand assembly for metal-organic frameworks as good candidates.

For at least one of these models we want ...

• The weights in a public model repository (ideally in Hugging Face if the weights aren't too big)
• All the code and materials needed to run it on a Delta node. That includes ...
  • Job submission script
  • Path of conda env and list of all HPC modules needed (or equivalent environment setup info)
  • Test data to run the model against
  • Actual code to set up and run the model

This can be delivered in a new repo we set up in the Garden-AI org for this purpose.

Assumptions:

1. 2.

Acceptance Criteria

Given I have an allocation on Delta, when I want to manually run the HPC exemplar model, then I go to the HPC exemplar repo and I have all the materials I need to run the model.

[WillEngler]

Eliu suggests starting with this model:

Repo: https://gitlab.com/hyunp2/ai4molcryst_argonne Google Colab: https://colab.research.google.com/drive/1viGlpUIGmRW-kzqERj4fGzX9udEXVrXw?usp=sharing