nvwb converter

A bash script to turn an arbitrary git repo into an NVIDIA AI Workbench project.

usage

• open Jupyterlab from AI Workbench
• navigate to code/nvwb_converter.bash
• open a terminal in Jupyterlab (Click "+", then select Terminal)
• run bash nvwb_converter.bash
• when prompted, enter the Github URL for the TARGET repo to be converted into an nvwb project
  • tip: ensure you have write access to ensure changes can be pushed upstream, eg. create a fork if necessary
• when prompted, enter the Github URL for an existing AI Workbench project you would like to use as a TEMPLATE
• let the script run to completion

what it does

• transfers the environment config files from the template repo to the target repo
  • .project/configpacks
  • .project/spec.yaml
  • apt.txt
  • requirements.txt
  • preBuild.bash
  • postBuild.bash
  • variables.env
• notes if it overwrites any files in the target git repo
• updates the spec.yaml transferred from the template repo
  • updates the meta.name and meta.image fields
  • reads the new repo directory structure and updates the layout fields
  • resets mounts, secrets, and user-added applications to defaults
• finds and updates any docker-compose.yaml files transferred from the template repo
  • Adds the NVWB_TRIM_PREFIX environment variable to all detected services
• commits changes to the target repo and adds detailed commit message.

next steps

• once converted, push the committed changes of your target repo to upstream
• in AI Workbench, clone the project using the target Github URL with these newly pushed updates
• once built, you can now make necessary project-specific adjustments
  • configure any project secrets and API keys
  • configure project-specific environment variables
  • create custom project mounts
  • add any custom applications
  • update the docker-compose file