v0.1.5

Primary changes

This version of EQUINE updates the base, GP, and Protonet Equine models to be able to record feature_names and label_names, which are important when using the webapp for prediction visualization. The tests also check that the OOD scores remain the same after saving and loading a model. The example notebooks are updated to show how to use the new features.

What's Changed

• Bump pypa/gh-action-pypi-publish from 1.9.0 to 1.10.0 by @dependabot in #65
• Bump pypa/gh-action-pypi-publish from 1.10.0 to 1.10.1 by @dependabot in #67
• Update numpy requirement from <=2.1.0,>=1.22.0 to >=1.22.0,<=2.1.1 by @dependabot in #66
• Update hypothesis requirement from <6.112.0,>=6.105.0 to >=6.105.0,<6.113.0 by @dependabot in #68
• Bump pypa/gh-action-pypi-publish from 1.10.1 to 1.10.2 by @dependabot in #70
• Feature label names by @harryli0088 in #69
• added vis_support=true to GP train_model in toy example notebook by @harryli0088 in #71

New Contributors

• @harryli0088 made their first contribution in #69

Full Changelog: v0.1.4...v0.1.5

v0.1.5rc1

What's Changed

• Bump pypa/gh-action-pypi-publish from 1.9.0 to 1.10.0 by @dependabot in #65
• Bump pypa/gh-action-pypi-publish from 1.10.0 to 1.10.1 by @dependabot in #67
• Update numpy requirement from <=2.1.0,>=1.22.0 to >=1.22.0,<=2.1.1 by @dependabot in #66
• Update hypothesis requirement from <6.112.0,>=6.105.0 to >=6.105.0,<6.113.0 by @dependabot in #68
• Bump pypa/gh-action-pypi-publish from 1.10.1 to 1.10.2 by @dependabot in #70
• Feature label names by @harryli0088 in #69

New Contributors

• @harryli0088 made their first contribution in #69

Full Changelog: v0.1.4...v0.1.5rc1

v0.1.4

What's Changed

• Adding testing support for python 3.12 by @nukularrr in #52
• Removed timing deadline on highly variable test by @nukularrr in #53
• Protonet fixes by @RoundOffError in #61
• Other enhancements for typehints and version dependencies
• Ensured protonet backend works on the GPU by @stevenjson in #64

Full Changelog: v0.1.3...v0.1.4

v0.1.4rc1

What's Changed

• Adding testing support for python 3.12 by @nukularrr in #52
• Removed timing deadline on highly variable test by @nukularrr in #53
• Protonet fixes by @RoundOffError in #61
• Other enhancements for typehints and version dependencies

Full Changelog: v0.1.3...v0.1.4rc1

v0.1.3

What's Changed

*Enable concomitant visualization capability for the GP in the web-app by @nukularrr in #36
*Removed no_grad statement from GP forward method and moved model to device in gp by @gtbotkin in #42
*Various dependency updates in github actions, hypothesis, etc.

New Contributors

• @gtbotkin made their first contribution in #42

Full Changelog: v0.1.2...v0.1.3

v0.1.3rc

What's Changed

• Enable concomitant visualization capability for the GP in the web-app by @nukularrr in #36
• Removed no_grad statement from GP forward method and moved model to device in gp by @gtbotkin in #42
• Various dependency updates in github actions, hypothesis, etc.

New Contributors

• @gtbotkin made their first contribution in #42

Full Changelog: v0.1.2...v0.1.3rc

EQUI(NE)^2 (equine): Establishing Quantified Uncertainty for Neural Networks

The goal of this package is to make it simple to add modern uncertainty quantification (UQ) techniques to existing PyTorch models to produce label predictions with calibrated probabilities and out-of-distribution indicators.

What's Changed

Overall: two minor bugfixes, testing coverage increased, improved type hinting, and more comprehensive CI tools and actions.

• Feature/beartype by @nukularrr in #21
• Stevenjson/patch 1 by @nukularrr in #22
• Distance update by @RoundOffError in #28
• Auto-updated dependencies with dependabot

Full Changelog: v0.1.1...v0.1.2

v0.1.1

EQUI(NE)^2 (equine): Establishing Quantified Uncertainty for Neural Networks

The goal of this package is to make it simple to add modern uncertainty quantification (UQ) techniques to existing PyTorch models to produce label predictions with calibrated probabilities and out-of-distribution indicators.

v0.1.1rc5

EQUI(NE)^2 (equine): Establishing Quantified Uncertainty for Neural Networks

The goal of this package is to make it simple to add modern uncertainty quantification (UQ) techniques to existing PyTorch models to produce label predictions with calibrated probabilities and out-of-distribution indicators.

What's Changed

• Added the Zenodo-linked DOI via GitHub integration.
• Added automatic versioning via setuptools_scm

Full Changelog: v0.1.1rc4...v0.1.1rc5

v0.1.1rc4

Minor changes to add Zenodo DOI