Tutorial 1C: Auto-generate an AutoDoc for a Scikit-learn model

Overview

This tutorial walks through the process of how you can generate an AutoDoc for a model built in Scikit-learn. To generate an AutoDoc for a built model in Scikit-learn:

• Step 1: Download Scikit-learn model
• Step 2: Define the settings for the AutoDoc

Prerequisites

• Knowledge of Scikit-learn

Step 1: Download Scikit-learn model

To build an AutoDoc for a supervised learning model built in Scikit-learn, you need to download its model first.

note

• As a requirement, H2O AutoDoc requires the Scikit-learn model to be in a Pickle file format (preferably in a Joblib format)
  • To learn more, see Model persistence: Python specific serialization

Step 2: Define the settings for the AutoDoc

After downloading the Scikit-learn model, you need to define the settings for the AutoDoc.

1. In H2O AutoDoc, click + Create new AutoDoc.
2. In the New report list, select From Scikit model.
3. In the Report name box, enter a name for the AutoDoc (for example, Scikit-learn AutoDoc).
4. Click Next: Upload model.
5. To upload your Scikit-learn model, click Browse....
   
6. After uploading you model, click Upload Scikit model.
7. Click Next: Upload training & validation data.
8. Click Upload training data.
   Note

   Anytime you are preparing the settings for an AutoDoc for a built model in Scikit-learn, you need to upload the training dataset used to build the model.

9. Click Browse....
10. After uploading the train dataset, click Upload training data.
11. In the Select the target column used while training list, select the model's target column.
12. Click Next: Upload test data.
13. Click Skip test data.
    Note

    • For purposes of this tutorial, we will not upload the test dataset of the built model in Scikit-learn
    • Anytime you want to generate an AutoDoc for a built model in Scikit-learn, you don't need to provide a test or used test dataset
    • Not providing a test dataset will lead to the AutoDoc (report) not containing an overview of the validation dataset
14. Click Create AutoDoc.

Summary

This tutorial taught us how to auto-generate an AutoDoc for a Scikit-learn model.

---

Feedback

• Submit and view feedback for this page
• Send feedback about H2O AutoDoc to cloud-feedback@h2o.ai