After creating our container, it's time to run it.
Our first container didn't work (of course). After a few iterations we got it running. 🎉
After a container path whoopsie, the main challenge was fetching the active machine config from within the container.
Previously, we'd copy the notebook id into the benchmark then use the notebook API to fetch the machine config. We found the API to describe the custom job by ID which has the machine info … but you don't have an ID until the job is created. 🤨
This StackOverflow answer had the key. You do get to set a display name, then you can fetch all jobs filtered on that name.
So we'll need to make sure those job names (nominally for display) are actually unique identifiers…
Top comments (0)