-
Book Overview & Buying
-
Table Of Contents
Practical C# Projects with .NET
By :
If you ran this chapter's code, you likely noticed that training the model took a significantly longer amount of time than using that model to generate predictions.
Model training is inherently a slower activity, particularly with AutoML, which must train and compare many different models and their hyperparameter values.
Additionally, model training usually involves human involvement as model metrics and characteristics are reviewed and approved prior to a model being deployed to production.
Conversely, generating predictions is usually nearly instantaneous, where new values are typically inserted into a mathematical formula that then produces a specific result representing the prediction.
As a result, we don't typically train a new model every time we want to generate predictions. Instead, we train a model and save it to disk so it can be loaded later.
With ML.NET, we can use our MLContext object's Save method to save the...