Summary

The key takeaway from this chapter should be that any machine learning model can be deployed like any other piece of code. The only difference is that we have to make room for being able to load the model again from disk. To do this, first, we need to train a model and write the model code and weights to disk using joblib. Then, we need to build a predict function, which is separated from training. Finally, we expose what we have done by using Flash with Jinja2 HTML templates.