Papermill

It is possible to run a Jupyter notebook as a Flyte task using papermill. Papermill executes the notebook as a whole, so before using this plugin, it is essential to construct your notebook as recommended by papermill. When using this plugin, there are a few important things to keep in mind:

1. This plugin can be used for any task - type.

   • It can be python code, which can be a tensorflow model, a data transformation, etc - but things that run in a container and you would typically write in a @task.
   • It can be a flytekit.dynamic workflow.
   • It can be a any other plugin like Spark, SageMaker etc, ensure that the plugin is installed as well

2. Flytekit will execute the notebook and capture the output notebook as an .ipynb file and an HTML rendered notebook as well
3. Flytekit will pass the inputs into the notebook as long as you have the first cell annotated as parameters and inputs are specified
4. Flytekit will read the outputs from the notebook, as long as you use annotate the notebook with outputs and outputs are specified

Installation

To use the flytekit papermill plugin simply run the following:

$ pip install flytekitplugins-papermill