v1.2.0

What's Changed

• Save model weights as safetensors by default, load either format by @lfoppiano in #245
• Configure Dependabot for minor updates and GitHub Actions (dev) by @lfoppiano in #247
• chore(deps): bump the actions group with 2 updates by @dependabot[bot] in #249
• fix: tag with the ChainCRF models, whose decoders return a tensor by @lfoppiano in #241
• chore(deps): bump the documentation group across 2 directories with 3 updates by @dependabot[bot] in #251
• chore(deps): bump pandas from 2.2.3 to 2.3.3 by @dependabot[bot] in #252
• chore(deps): bump lmdb from 2.1.1 to 2.3.0 by @dependabot[bot] in #254
• chore(deps-dev): update ruff requirement from <0.16.0 to <0.17.0 by @dependabot[bot] in #253
• ci: build the branches concurrently by @lfoppiano in #255
• chore(deps): bump transformers from 5.7.0 to 5.17.0 in the huggingface group across 1 directory by @dependabot[bot] in #250
• Give each word the label of its first sub-token when tagging with a transformer by @lfoppiano in #242
• feat: add --suffix to keep several models of a task and architecture by @lfoppiano in #246
• warn when the sequence is larger than the model max_tokens_length by @lfoppiano in #244
• Get models from huggingface by @lfoppiano in #258
• Add a warning when --transformer and --embeddins are ignored by @lfoppiano in #260
• Publish models to huggingface by @lfoppiano in #259
• Integration/fixes and features by @lfoppiano in #261
• Improve model loading from URLs and directories by @lfoppiano in #262
• feat: implement sliding windows for larger sequences by @lfoppiano in #263
• feat: read the texts of a classification data set from disk by @lfoppiano in #264
• feat: report the mean and the standard deviation of the scores over t… by @lfoppiano in #266
• Fix regression in saving safetensors by @lfoppiano in #269
• chore(deps): bump unidecode from 1.3.2 to 1.4.0 by @dependabot[bot] in #270
• chore(deps): bump scikit-learn from 1.6.1 to 1.7.2 by @dependabot[bot] in #271
• Add modern static embeddings by @lfoppiano in #238
• Add --seed in every application as CLI parameter by @lfoppiano in #273
• Contextual embeddings from a frozen transformer by @lfoppiano in #239
• fix: fit the preprocessor of a fold model on the training data of the… by @lfoppiano in #265
• feat: features channel for text classification by @lfoppiano in #268
• Rewrite the CRF Loss computation to increase GPU usage during training by @lfoppiano in #275
• Fixes to the contextual embeddings by @lfoppiano in #277
• Test array by @lfoppiano in #278
• Stack embeddings, e.g. static and contextual ones, by concatenation by @lfoppiano in #257
• chore(deps): bump transformers from 5.17.0 to 5.18.0 in the huggingface group across 1 directory by @dependabot[bot] in #279
• Add an nfold profile to the array submitter by @lfoppiano in #280
• audit before release by @lfoppiano in #285

Full Changelog: v1.1.0...v1.2.0