probabl-ai/skore

🎯 Why Skore?

When it comes to data science, you have excellent tools at your disposal: pandas and polars for data exploration, skrub for stateful transformations, and scikit-learn for model training and evaluation. These libraries are designed to be generic and accommodate a wide range of use cases.

But here's the challenge: Your experience is key to choosing the right building blocks and methodologies. You often spend significant time navigating documentation, writing boilerplate code for common evaluations, and struggling to maintain clear project structure.

Skore is the conductor that transforms your data science pipeline into structured, meaningful artifacts. It reduces the time you spend on documentation navigation, eliminates boilerplate code, and guides you toward the right methodological information to answer your questions.

What Skore does for you:

• Structures your experiments: Automatically generates the insights that matter for your use case
• Reduces boilerplate: One line of code gives you comprehensive model evaluation
• Guides your decisions: Built-in methodological warnings help you avoid common pitfalls
• Maintains clarity: Structured project organization makes your work easier to understand and maintain

⭐ Support us with a star and spread the word - it means a lot! ⭐

🧩 What is Skore?

The core mission of Skore is to turn uneven ML development into structured, effective decision-making. It consists of two complementary components:

• Skore Lib: the open-source Python library (described here!) that provides the structured artifacts and methodological guidance for your data science experiments.
• Skore Hub: the collaborative platform where teams can share, compare, and build upon each other's structured experiments. Learn more on our product page.

⚡️ Quick start

Installation

With pip

We recommend using a virtual environment (venv). You need python>=3.11.4.

Then, you can install skore by using pip:

# If you plan to use Skore locally
pip install -U skore
# If you wish to interact with Skore Hub as well
pip install -U skore[hub]
# If you wish to log projects to MLflow
pip install -U skore[mlflow]

With conda

skore is available in conda-forge both for local and hub use:

conda install conda-forge::skore

You can find information on the latest version here.

Get structured insights from your ML pipeline

Evaluate your model and get comprehensive insights in one line:

from sklearn.datasets import make_classification
from sklearn.linear_model import LogisticRegression
from skore import evaluate

X, y = make_classification(n_classes=2, n_samples=100_000, n_informative=4)
clf = LogisticRegression()

# Get structured insights that matter for your use case
cv_report = evaluate(clf, X, y, splitter=5)

# See what insights are available
cv_report.help()

# Example: Access the metrics summary
metrics_summary = cv_report.metrics.summarize().frame()

# Example: Get the ROC curve
roc_plot = cv_report.metrics.roc()
roc_plot.plot()

Learn more in our documentation.

🛠️ Contributing

Join our mission to promote open-source and make machine learning development more robust and effective. If you'd like to contribute, please check the contributing guidelines here.

👋 Feedback & Community

• Join our Discord to share ideas or get support.
• Request a feature or report a bug via GitHub Issues.

Support

Skore is tested on Linux and Windows, for at most 4 versions of Python, and at most 4 versions of scikit-learn:

• Python 3.11
  • scikit-learn 1.6
  • scikit-learn 1.9
• Python 3.12
  • scikit-learn 1.6
  • scikit-learn 1.9
• Python 3.13
  • scikit-learn 1.6
  • scikit-learn 1.9
• Python 3.14
  • scikit-learn 1.7
  • scikit-learn 1.8
  • scikit-learn 1.9

---

Brought to you by