neptune-contrib: open-source contributions to Neptune.ml

This library is a collection of helpers and extensions that make working with Neptune app more effective and better. It is build on top of neptune-client and gives you option to do things like:

• interactive visualizations of experiment runs or hyperparameters
• running hyper parameter sweeps in scikit-optimize, hyperopt or any other tool you like
• monitor training of the lightGBM or fastai models with a single callback
• much more

Enjoy the following integrations:

And the best thing is you can extend it yourself or… tell us to do it for you :).

Get Started

• Installation
• Overview

Examples

• Data versioning
• Interactive experiment run comparison
• Image directory snapshoting
• Hyper parameter comparison
• Log binary classification metrics
• Log fairness classification metrics
• Integrate with Sacred
• Monitor lightGBM training
• Monitor fast.ai training
• Visualize project progress
• Sync experiments with Neptune via json file
• Telegram bot

User Guide

• api.utils
• hpo.utils
• bots.telegram_bot
• monitoring.lightgbm
• monitoring.fastai
• monitoring.metrics
• monitoring.fairness
• monitoring.optuna
• monitoring.sacred
• monitoring.skopt
• monitoring.utils
• sync.with_json
• versioning.data
• viz.experiments
• viz.projects

Bug Reports and Questions

neptune-contrib is MIT-licensed and the source code is available on GitHub. If you find yourself in any trouble drop an isse on Git Issues, fire a feature request on Git Feature Request or ask us on the Neptune community forum or Neptune community spectrum.

Contribute

We keep an updated list of open issues/feature ideas on github project page Github projects. If you feel like taking a shot at one of those do go for it! In case of any trouble please talk to us on the Neptune community spectrum.

Indices and tables

• Index
• Module Index