whylogs

Stale
GitHub Jupyter Notebook Apache-2.0

Description

Lightweight data and model profiling library that captures feature statistics for drift detection.

Key Features

  • Lightweight profiling — One line to produce data and model feature stats
  • Privacy-safe — Profiles are statistics, not raw data
  • MLflow integration — Store profiles as MLflow artifacts
  • Drift detection — Combine with WhyLabs / Evidently for data drift
  • Multimodal — Tabular, text, image, and embedding supported
  • Embeddable — Python library plus container image

Use Cases

💡 Capture input/output distributions for LLM drift monitoring.
💡 Record per-batch feature stats in ML training pipelines.
💡 Upload profiles to WhyLabs for dashboard visualization.

Quick Start

# Install
pip install whylogs
# Profile a DataFrame
import whylogs as why
import pandas as pd
df = pd.read_csv('data.csv')
result = why.log(df)
result.view()

Related Projects