Skip to content

Latest commit

 

History

History
40 lines (34 loc) · 2.06 KB

template_developers.md

File metadata and controls

40 lines (34 loc) · 2.06 KB
Raw

Template Developers Manual

The easiest way to add a new template to MLT is to make a copy of one of the existing templates and then modify it. The following instructions will go through those steps:

  1. Start by making a copy of one of the folders in the mlt-templates directory. Either start with something simple like the hello-world template, or pick a template that is similar to the one that you are going to add.

  2. Rename the new folder with the name for your template.

  3. Update the README.md file with a short description of your template. Note that this is the description that will be displayed with mlt templates list

  4. Other files that typically need to be modified for new templates are:

  • *.py files
  • requirements.txt - updated with any libraries that your app uses
  • crd-requirements.txt - (optional) updated with any operators your model depends on. By default crd-check is performed during the execution of both mlt init and mlt deploy. The crd-check command compares crds specified in crd-requirements.txt against crds in your cluster. Example
  • Dockerfile - modified if the name of the python file to execute is different, etc.
  • k8s-templates/job.yaml - required for the kubernetes job, may need to be modified if a CRD is used, if multiple replicas are needed, or if environment variables need to be set, etc. Note that all files in the k8s-templates directory can have variables like $app, $run, and $image that will be subbed into the template when the app is deployed.
  • parameters.json - (optional) used to define any additional parameters specific to your template. These parameters are added to the app's mlt.json file when it's initialized. The parameter values are subbed in to the k8s-templates directory when the app is deployed.
  1. To test your template, you need to use the --template-repo parameter with mlt templates list and mlt init in order to specify the git url for template repo (otherwise, this defaults to https://github.com/IntelAI/mlt).