Use it when
- You want a lightweight tool to cover data versioning, model registry, experiment, and artifact tracking.
- You want to record and monitor model training, evaluation, or production runs.
- You want a tool that covers a wide range of metadata types to log and display, from model parameters to audio and video files.
- You want 25+ built-in integrations with popular Python libraries.
Watch out
- Only the Neptune client is open-source and requires signing up.
- The Neptune client covers data logging to the Neptune server, data queries, and downloads of your logged data.
- The free plan is only for individual use up to 200 monitoring hours/month.
- Self-hosted deployment requires an Organization plan.
Example stacks
Airflow + MLflow stack
Installation
pip install neptune-client