Take skore for a test drive!

Intro

This repository is here to help you test and provide feedback on skore, a new Python library that aims to enhance scikit-learn for data scientists. skore is currently a work-in-progress, and more features are to be expected.

Skore is developed by the product team at Probabl, a scikit-learn company. For more info, check us out: Notion page.

Agenda

In total, this test drive should last under an hour.

1. Presentation of Probabl and skore: ~5 min
2. Environment setup: ~5 min
3. User testing tasks: ~35 min
   • Here is a notebook to get you started: compare_models.ipynb.
4. Feedback Q&A: ~15 min

Set-up: create your virtual environment

1. Clone this GitHub repository locally.
2. Create a Python virtual environment with your usual method (conda, virtualenv...), with python>=3.9.
3. Install the libraries in requirements.txt with (activate your venv beforehand):

   pip install -r requirements.txt

Goal of the test

The goal is to build a couple of models, and to compare them, using skore when relevant.

Resources:

• Notebook: compare_models.ipynb
• Skore documentation

Your feedback

Let us know what you think of skore! Here are some topics on which we would appreciate your feedback:

1. Overall experience
   • First impressions?
   • Pain points?
   • Unexpected discoveries?
2. Analysis flow
   • How natural was the storage/retrieval?
   • Dashboard creation experience?
   • What would improve your workflow?
3. Suggestions
   • Missing features?
   • Onboarding improvements?
   • Would you use skore in your day-to-day work? Would you recommend skore to a colleague?

You can share your feedback:

• either directly by mail to the Product Team,
• or on the #skore channel of our Discord server.

How to help?

• Give a star to the skore repository
• Tell your friends, invite them to test skore with this repo (https://github.com/probabl-ai/skore-user-test)
• Bring discussions to our Discord server