`flippers` is a weak supervision library for creating high quality labels using your domain kownledge and weak supervision sources.
Project description
flippers - A Weak Supervision Library
(in construction)
flippers
is a Python library for weak supervision, which allows you to leverage your domain knowledge, heuristics and other weak supervision sources to generate high-quality labels for your training data.
Features
flippers
includes a number of features for weak supervision, including:
- Simple tools to analyse your labeling functions,
- Multiple label models including a from-scratch implementation of the label model used in the
snorkel
library and featuring enhanced ways to predict probabilities, - An extensive documentation with tutorials and an API reference.
Installation
To install the latest version of flippers
, simply run:
pip install flippers
Quick Start
Documentation
To quickly get started with flippers
, you can begin by exploring the documentation and running through the examples provided. The examples cover a variety of use cases and techniques, which can help you to get a feel for how to apply flippers
to your own projects.
Example
- Analyzing your labeling functions:
analysis = flippers.analyis(L_train)
- Training a Label Model and doing inference:
label_model = flippers.models.SnorkelModel(polarities, class_balances)
label_model.fit(L_train)
label_model.predict_proba(L)
Discussion
Troubleshooting
If you have any questions or issues with flippers
, please consult the documentation or reach out in the GitHub issues page for support.
Contributing
flippers
is an open-source project, and contributions are welcome! If you would like to contribute to flippers
, please read the CONTRIBUTING.md
file for guidelines and instructions.
Related projects
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for flippers-0.0.1b3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 95b2d82381f5377b971211b4feb2c93708b1382b26f2c0db9bb37fa9ad57f737 |
|
MD5 | b381b32e2fd9db6f7a0efb4d0aff17ed |
|
BLAKE2b-256 | 9a28b61d975cdf744db51e442af5f71dccc4240a092195756fe0d170e17dca37 |