## What's Changed * prepare documentation deployment by @Ingvarstep in https://github.com/urchade/GLiNER/pull/310 * Fix: move apex.amp import to avoid crash when importing GLiNER by @tannonk in https://github.com/urchade/GLiNER/pull/313 * Fix ONNX model loading when torch file is absent by @Ingvarstep in https://github.com/urchade/GLiNER/pull/317 * loss fc fixed by @BioMikeUkr in https://github.com/urchade/GLiNER/pull/322 * Introduce new token-level architectures by @Ingvarstep in https://github.com/urchade/GLiNER/pull/321 * Add handling of empty strings by @Ingvarstep in https://github.com/urchade/GLiNER/pull/325 * fix(onnx): use explicit max_width in SpanMLP.view() for ONNX export by @ajroetker in https://github.com/urchade/GLiNER/pull/318 * Fix training issues and compatability with transformers v4.x and v5 by @Ingvarstep in https://github.com/urchade/GLiNER/pull/326 * Feature/candidate labels by @Ingvarstep in https://github.com/urchade/GLiNER/pull/328 * Update README.md by @m-newhauser in https://github.com/urchade/GLiNER/pull/330 ## New Contributors * @tannonk made their first contribution in https://github.com/urchade/GLiNER/pull/313 * @ajroetker made their first contribution in https://github.com/urchade/GLiNER/pull/318 * @m-newhauser made their first contribution in https://github.com/urchade/GLiNER/pull/330 **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.24...v0.2.25

## What's Changed * fix masking interface for proper training span models by @Ingvarstep in https://github.com/urchade/GLiNER/pull/296 * Add inference-time sequence packing support by @vivekkalyanarangan30 in https://github.com/urchade/GLiNER/pull/292 * resolved packing params not propagating; added support for other backbones by @vivekkalyanarangan30 in https://github.com/urchade/GLiNER/pull/301 * Refactoring GLiNER, making it more modular and easier to use by @Ingvarstep in https://github.com/urchade/GLiNER/pull/305 * update transformer version by @urchade in https://github.com/urchade/GLiNER/pull/306 ## New Contributors * @vivekkalyanarangan30 made their first contribution in https://github.com/urchade/GLiNER/pull/292 **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.22...v0.2.23

## What's Changed * jieba3 added by @oleksandrlukashov in https://github.com/urchade/GLiNER/pull/269 * Update requirements.txt to not use onnxruntime-gpu for mac by @anny-lanai in https://github.com/urchade/GLiNER/pull/271 * Add StanzaWordsSplitter for language-aware tokenization by @anthonyyazdani in https://github.com/urchade/GLiNER/pull/276 * Implement Assymetric loss by @Ingvarstep in https://github.com/urchade/GLiNER/pull/272 * Fixed ALL_LAYERNORM_LAYERS bug by @dcrey7 in https://github.com/urchade/GLiNER/pull/278 * Add generation with constraints by @Ingvarstep in https://github.com/urchade/GLiNER/pull/283 * loss reduction argument fixed by @BioMikeUkr in https://github.com/urchade/GLiNER/pull/286 * Labels decoder by @Ingvarstep in https://github.com/urchade/GLiNER/pull/282 * Fix Cython modules usage by @Ingvarstep in https://github.com/urchade/GLiNER/pull/287 * Converts `texts` given as a single str to list by @jackboyla in https://github.com/urchade/GLiNER/pull/288 * fix SpanBiEncoderProcessor label assignment by @cceyda in https://github.com/urchade/GLiNER/pull/289 ## New Contributors * @anny-lanai made their first contribution in https://github.com/urchade/GLiNER/pull/271 * @anthonyyazdani made their first contribution in https://github.com/urchade/GLiNER/pull/276 * @dcrey7 made their first contribution in https://github.com/urchade/GLiNER/pull/278 * @BioMikeUkr made their first contribution in https://github.com/urchade/GLiNER/pull/286 * @jackboyla made their first contribution in https://github.com/urchade/GLiNER/pull/288 * @cceyda made their first contribution in https://github.com/urchade/GLiNER/pull/289 **Full Changelog**: https://github.com/urchade/GLiNER/compare/0.2.21...v0.2.22

## What's Changed * Update relation_extraction.py by @rafmacalaba in https://github.com/urchade/GLiNER/pull/255 * add cache_dir parameter to AutoConfig, AutoTokenizer by @bit2244 in https://github.com/urchade/GLiNER/pull/250 * add load from config by @urchade in https://github.com/urchade/GLiNER/pull/257 * Add support of Flash DeBERTa by @Ingvarstep in https://github.com/urchade/GLiNER/pull/259 ## New Contributors * @rafmacalaba made their first contribution in https://github.com/urchade/GLiNER/pull/255 * @bit2244 made their first contribution in https://github.com/urchade/GLiNER/pull/250 **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.19...v0.2.20

## What's Changed * docs: update README_Extended.md by @eltociear in https://github.com/urchade/GLiNER/pull/239 * add neg sampling by @urchade in https://github.com/urchade/GLiNER/pull/251 * modify trainer by @urchade in https://github.com/urchade/GLiNER/pull/252 * new v by @urchade in https://github.com/urchade/GLiNER/pull/253 * add data_type in config by @urchade in https://github.com/urchade/GLiNER/pull/254 **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.17...v0.2.19

**best support of multi-label for span-based ner training **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.15...v0.2.16

## What's Changed * make `onnxruntime-gpu` optional by @moritzwilksch in https://github.com/urchade/GLiNER/pull/223 **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.14...v0.2.15

## What's Changed * Bug fix: Tensors should not be moved to a device when using onnx by @azikoss in https://github.com/urchade/GLiNER/pull/196 * Bug fix: Tensors should not be moved to a device when using onnx (in evaluate function) by @wojiaodawei in https://github.com/urchade/GLiNER/pull/199 * Change how the config file is loaded by @eriknovak in https://github.com/urchade/GLiNER/pull/206 * Add GPU Support for ONNX Models by @KarimLulu in https://github.com/urchade/GLiNER/pull/201 * Simplify inference and evaluation for multi-task models by @Ingvarstep in https://github.com/urchade/GLiNER/pull/217 ## New Contributors * @azikoss made their first contribution in https://github.com/urchade/GLiNER/pull/196 * @wojiaodawei made their first contribution in https://github.com/urchade/GLiNER/pull/199 * @eriknovak made their first contribution in https://github.com/urchade/GLiNER/pull/206 * @KarimLulu made their first contribution in https://github.com/urchade/GLiNER/pull/201 **Full Changelog**: https://github.com/urchade/GLiNER/compare/v0.2.13...v0.2.14