Nodes
Waiting
Finished recently
Each job gets all 8 GPUs of one node, runs in its own container, and starts on whichever node frees up first.
To queue one from a cluster node, run gitm-job submit --name my-run --model org/model
and check your place with gitm-job ls.
Without a shell on the cluster, submit through the Slurm REST API with your Slurm token.