Stratosphere
A lightweight experimentation toolkit for data scientists.
Designed to be accessibile, fast and robust. stratosphere lets you:
- Define programmatically your experiments
- Execute them in parallel with different backends
- Track their predictions and metrics
- Store them as plain tables and pickled objects
- Query them with SQL, Pandas or Python
How is it possible?
- Fast and light: Try it now with JupyterLite, it runs entirely in the browser
- Built on top of solid components: SQLAlchemy, SQLite, Pandas, Joblib and Dask
- Designed for data accessibility: Use it where you need it. For example, you can experiment in-memory, persist locally, replicate tables on render.com, and publish results on preset.io (all free tiers!)
Installation
On Python >= 3.8.0:
- With PyPI:
pip install "stratosphere[complete]" --upgrade# Install everything - With Poetry:
poetry add stratosphere@latest --extras complete# Install everything
On Python 3.7.13-15 with Google Colab or
Binder, execute the following at the beginning of the notebook:
# Install dependencies
!pip install pandas joblib sqlalchemy sqlalchemy-utils ulid-py cloudpickle colorama tqdm --upgrade
# Install [complete] extras, required to run the tutorial
!pip install tabulate scikit-learn dask[complete] --upgrade
# Install the latest stratosphere version, ignoring the python version and dependencies
!pip install stratosphere --upgrade --ignore-requires-python --no-dependencies
# Resolve ContextualVersionConflict errors, due to broken dependencies
import importlib
import pkg_resources
importlib.reload(pkg_resources)
You can also install only the stratosphere library (dropping the [complete] extras).
Modules like stratosphere.utils.dask, stratosphere.utils.metrics, and stratosphere.utils.widgets
won't work until you also install dask[complete], scikit-learn, and ipywidgets, respectively.
These extras are required to run some of the tutorial notebooks.
Documentation
- In the intro notebook, you'll be exposed to the key concepts. It runs entirely in the browser.
- In the tutorial notebooks, you'll cover the advanced topics. Working locally, on Colab and on Binder.
Project pages
License
This project is licensed under the terms of the BSD 3-Clause License.
Development
See the development page.
Contributing
Work in progress!
