This is an IBM ppc64le physical GPU server in our local datacenter.
Before running docker-compose in this directory, please be sure to source the dm:
[sofwerx::] icbmbp:ibm-minsky ianblenke$ swx dm env ibm-minsky-0
[sofwerx:ibm-minsky:ibm-minsky] icbmbp:ibm-minsky ianblenke$
Due to DOCKER_COMPOSE=ibm-minsky.yml in the ibm-minsky environment, the ibm-minsky.yml here is the docker-compose.yml that is used when a docker-compose is run.
This docker-compose will spin up a traefik SSL reverse proxy that will allocate an SSL cert automatically from Let's Encrypt using ACME.
This is configured to run tensorflow using GPU, and the 500ms periodicity for triangulate reflects this.
docker-machine create -d generic --generic-ip-address 172.109.152.115 --generic-ssh-port 9022 --generic-ssh-key ${devops}/secrets/ssh/sofwerx --generic-ssh-user swxadmin --generic-engine-port 9376 --engine-storage-driver overlay2 swx-u-ub-minsky
swx dm import swx-u-ub-minsky