Is there any documentation about how the AI Workbench project repository is structured?
I’m talking about things like code/data/models, along with .project/spec-yaml, pre/post build, etc.
I think we’re doing OK fumbling through various aspects, and borrowing from random repos or just setting up a new project, but my gut says there is probably something out there that already exists as a starting point? And/or, is there documentation that explains what all the things are and how they are used?
I know we can do it through AI Workbench, but there’s quite a bit I can do outside of NVWB and I’d rather work locally until I have a reason to clone.
Please describe your issue or request: (tick the boxes after creating this topic):
Please tick the appropriate box to help us categorize your post
Bug or Error
Feature Request
Documentation Issue
Other
Yes, there’s documentation. We want to improve it, so critical feedback is very valuable. See below for relevant docs.
Outside of the project specification and the required fields and locations for certain files (.project, requirements.txt, etc), the project structure is flexible and up to an end user.
For example, you aren’t tied to the code, data and models folder structure. You can change that as well as indicate the versioning backend as you wish.
However, you need to manually edit the spec.yaml file to account for these changes. We don’t yet have affordances in the UI or CLI for this aspect.
The project specification is also related to an image label specification that’s used for the base containers. You can see more on that here: Use Your Own Container - NVIDIA Docs