Please create a Colab notebook instead of a Kubeflow workflow or docker container

[SudoPseudo]

Why you need this feature:
I do not understand Kubeflow. I would like to try out the code for training and evaluating a model using synthetic data, but I am unable at the current time because of the code's complexity.

Describe the solution you'd like:
A colaboratory notebook, with simple code for downloading the synthetic and real dataset, training a model on the synthetic dataset, and then fine-tuning it on the real images.