Pennant

The GPU and flux GPU variants here are not automated in the repository - they were built manually (locally) and pushed to the same registry. We first built the gpu variant to try with the "add on the fly" flux view:

docker build -f Dockerfile.gpu -t ghcr.io/converged-computing/metric-pennant:gpu .
docker push ghcr.io/converged-computing/metric-pennant:gpu

We found that flux (hwloc) didn't detect the GPU, and we think it's because the view install with hwloc is separate from the cuda install. So we decided to try a base image that also has flux:

docker build -f Dockerfile.flux-gpu -t ghcr.io/converged-computing/metric-pennant:gpu .
docker push ghcr.io/converged-computing/metric-pennant:gpu