Q: At OLCF, is Singularity set up such that I can access GPUs from inside the container? And is there a cache/repo service for the facility (e.g. Shifter repo at NERSC)?
A: Yes, you can access the GPUs from the container. For container repo, you could use ORNL’s gitlab (it comes with one available to projects). We don’t have a container repo from the facility though.

Q: Can everyone access alphafold2 container on summit?
A: Yes, instructions are located here: https://github.com/olcf/AlphaFold_on_Summit

Q: Does a container need the same linux kernel to run? Can a container created on a Cray system that has its own Linux work on a non-Cray system with a stock Linux, even though both machines may have the same architecture?
A: See very end of recording (during recording Q&A) for answer

Q: What’s best practices for storing the SIF file? Home? Scratch?
A: Probably the NFS based project space - it has more space than $HOME and won’t get purged like scratch, so `/ccs/proj/<project-id>/`

Q: If I have an application that can only run on earlier CUDA version (<= cuda 8.0), can I build a container and run on summit or other HPC systems that have newer CUDA versions?
A: See very end of recording (during recording Q&A) for answer. Answer is essentially: it is guaranteed to work in that case but can be tested.