Testing

ML Testing is typically conducted in the research phase and is the first component in Continuous ML Validation.

Testing is an extremely important part of any software development process, and for various reasons hasn’t evolved as consistently and healthily as some other components in the ML tech stack. It’s time to change that.

Testing

Continuous
Validation

CI/CD

Monitoring

Root Cause Analysis

Why Testing?

Adopt Best Practices

While most software teams have clear methodologies around testing and QA, this is a mess for most teams of ML Practitioners.

Are you using ML testing best practices? Do you have standardized processes?

Standardize Research Reviews

Some teams don’t really do ML testing at all. Other teams have well-defined research review processes, usually involving tedious work that your top researchers do not have the time for!

Set Up Processes That Enable You to Scale

You will need rigorous testing to speed up model development and maintenance. Usually, that requires a lot of coding. The best way to enable scale is to use a testing framework covering all models with minimal coding efforts.

You Already Know You Should Start

Yes, you already know we’re right about this. You just never had a simple solution at your fingertips. Well, now you do. And it's free!

How Deepchecks Open Source Supports Testing

from deepchecks.tabular.suites import train_test_validationvalidation_suite = train_test_validation(...)validation_suite.run(train_ds, test_ds)

Some tools can assist with streamlining different aspects of validating ML. However, Deepchecks offers a one-stop shop for many different use cases. In most cases, this starts with testing, and the assets used for testing are later utilized for CI/CD and monitoring.

As a Python package, Deepchecks Open Source is highly customizable and integrates perfectly within Jupyter or Pycharm (no external interfaces or dashboard!).

Learn More

2.5K