v0.5.2

What's Changed

• [pre-commit.ci] pre-commit suggestions by @pre-commit-ci in #78
• Limit pytorch lightning version to prevent breaking changes by @dcferreira in #83
• Update mac version to latest in github workflow by @laurahanu in #89
• [pre-commit.ci] pre-commit suggestions by @pre-commit-ci in #88
• Bump transformers from 4.22.1 to 4.30.0 by @dependabot in #85
• fix for transformers 4.35.2 by @Vela-zz in #98
• Workflow to update pypi by @laurahanu in #102

New Contributors

• @dcferreira made their first contribution in #83
• @dependabot made their first contribution in #85
• @Vela-zz made their first contribution in #98

Full Changelog: v0.5.1...v0.5.2

Housekeeping updates & fixes

What's Changed

• Run tests workflow on a schedule by @jamt9000 in #51
• [pre-commit.ci] pre-commit suggestions by @pre-commit-ci in #47
• Map loaded checkpoint to specified device by @gregpriday in #54
• [pre-commit.ci] pre-commit suggestions by @pre-commit-ci in #55
• Parameters in evaluate.py by @anitavero in #57
• [pre-commit.ci] pre-commit suggestions by @pre-commit-ci in #67
• fix worker numbers type by @omidforoqi in #68
• Remove --use-feature=2020-resolver pip flag by @jamt9000 in #73
• Pin transformers version to 4.22.1 by @laurahanu in #76

New Contributors

• @pre-commit-ci made their first contribution in #47
• @gregpriday made their first contribution in #54
• @anitavero made their first contribution in #57
• @omidforoqi made their first contribution in #68

Full Changelog: v0.5.0...v0.5.1

Offline model loading and fixed transformers version

What's Changed

• freeze PL requirements by @Borda in #40
• allow Load offline, expose local path by @Borda in #39
• Pre-commit and import fixing by @Borda in #36
• Updated pytorch-lightning version by @laurahanu in #50

New Contributors

• @Borda made their first contribution in #40

Full Changelog: v0.4.0...v0.5.0

Updated multilingual model & consistent class names

• Updated the multilingual model weights used by Detoxify with a model trained on the translated data from the 2nd Jigsaw challenge (as well as the 1st). This model has also been trained to minimise bias and now returns the same categories as the unbiased model. New best AUC score on the test set: 92.11 (89.71 before).
• All detoxify models now return consistent class names (e.g. "identity_attack" replaces "identity_hate" in the original model to match the unbiased classes).

Updated weights for multilingual model

New improved weights for the multilingual Detoxify model trained on the translated data from the 2nd Jigsaw Challenge as well as from the 1st. Trained with same labels as the unbiased model.

v0.3.0. Updated unbiased model

• New improved unbiased model and updated data loaders to replicate
• tests to check torch.hub is loading the models
• script to convert saved weights to detoxify format

Updated weights for unbiased model

• New improved weights for the unbiased Detoxify model trained on the datasets provided by the 1st and 2nd challenges

Minor release to specify device when loading detoxify models

v0.2.2

fixes #13

Updated code with lightweight models

0.2.0

added news section to readme

Releasing lightweight models

Added lightweight checkpoints trained with Albert for original and unbiased models.