Hi simple question - what is the easiest way to provision / run the following project in Workbench?
I want to make it as accessible as as easy to load as possible in Workbench - for a group of people who have Workbench installed (acknowledging it may be easier to just have them run the lab in CoLab, but keen to encourage the use of Workbench if practical)
Specifically this lab
Andrew
Thanks for the post.
The “easiest” thing to do is to get the proper image labels on the container, post it to a public registry, and then create a project using it as the base image.
We have docs on this here, but they aren’t super useful…
I need to rewrite them, so working on this image would be a good exercise.
Anyways, at a high level, you just need to get into the container to figure out a few key items and then write a Dockerfile that pulls the image and adds that info in the appropriate format.
What’s your timeline?
Tyler
Thanks for the reply / Not time constrained / I’ve tried creating a simple repo to test a few things out but I keep getting a read error in WB / using clone, new project or import
The ideal would be have one link /
Got it.
You can create deep links out of your GitHub repos right now.
See here: Share AI Workbench Projects with Deep Links — NVIDIA AI Workbench User Guide
If the user has Workbench already installed, then the flow will push them to select a location to clone the project to.
Also, these RAPIDS base images ( RAPIDS Notebooks | NVIDIA NGC) are built with the workbench spec’d labels on them.
You can see the layers here: RAPIDS Notebooks | NVIDIA NGC
You should be able to create a project using any of these bases and the Custom Container option in the Project creation flow.
The custom base image you use should be public.
See the screen shots.
That’s super helpful / thank you
