Data science collaboration hub
The most lightweight experiment management tool that fits any workflow.
Use as a service or deploy on any cloud or your own hardware.
No credit card required. Takes 5 minutes to get started.
HOW IT WORKS
Neptune allows you to track the entire experimentation process
Everything is backed-up and organized, ready to be accessed, reproduced and shared with others.
Track and version your notebooks
Log all your notebooks directly from Jupyter or Jupyter Lab.
All you need is to install a Jupyter extension.
Manage your experimentation process
Neptune tracks your work with virtually no interference to the way you like to do it. Decide what is relevant to your project and start tracking:
- Metrics
- Hyperparameters
- Data versions
- Model files
- Images
- Source code
Integrate with your workflow easily
Neptune is a lightweight extension to your current workflow. Works with all common technologies in data science domain and integrates with other tools. It will take you 5 minutes to get started.
# Track your notebooks
pip install neptune-notebooks
jupyter nbextension enable ——py neptune-notebooks
# Track all experiment-related objects
import neptune
neptune.init()
neptune.log_metric(’auc’, score)
neptune.log_image(‘model_diagnostics’, ‘roc_auc.png’)
neptune.log_artifact(’model_weights.h5’, ’.’)
neptune.set_property(‘data_version’, sha1(data_train))
# Host MLflow or TensorBoard runs on Neptune
neptune@ubuntu:~$ neptune tensorboard path/to/logdir
neptune@ubuntu:~$ neptune mlflow path/to/project
Works with your favourite frameworks and tools
Compare notebooks like source code
Record your exploration process and analyze diffs between checkpoints. Select two notebooks and compare their content, code and outputs, side-by-side just like source code.
Version and compare experiments
Keep track of your progress and reproduce results easily. Neptune tracks all the details about every single experiment you run. You can tag, filter, sort and compare your experiments.
Collaborate easily
Discuss data science work in context, right where it is created. Explain your point by adding a comment to a particular object or simply share a link. Mention your teammates, spark discussions and discover insights – everything in a single place.
EXAMPLE pROJECTS
Hyperparameter Optimisation in Python
In a series of blog posts Jakub is comparing Python hyperparameter optimisation libraries: eg. Scikit-Optimize, Hyperopt, Optuna, hpbanster, BayesianOptimisation, Sherpa and SMAC3.
Binary classification
In this example project we walk you through data exploration and feature extraction all the way to tuning machine learning model. The whole process is described in the Wiki space.
Take advantage of
team collaboration.
seamless tracking.
organized work.
Start today!