v0.7.4

Changelog

Features 🏎️

• #577 - don't do any network calls if model is loaded from cache @joein
• #578 - expose enable_cpu_mem_arena onnx session option to handle onnxruntime memory allocation by @joein
• #582 - unlock hf hub and pillow versions by @joein
• #583 - add a new token_count method to text models by @joein @dancixx

Fixes 🕸️

• #585 - more granular numpy versioning

Thanks to everyone who contributed to this release @dancixx @joein and @tbung for all the reviews

v0.7.2

Changelog

Features 🏎️

• #542 - add MUVERA post processing module by @kacperlukawski @joein

Fixes 🕸️

• #547 - avoid adding padding token embeddings to the result in batched inference in late interaction models by @generall

Thanks to everyone who contributed to this release @kacperlukawski @generall @joein

v0.7.1

Changelog

Features 🏎️

• #532 - improved warnings for the recently changed models by @joein
• #522 - raise exceptions in case of incorrect pooling in custom models by @joein
• #521 - add property embedding_size and cls method get_embedding_size for easier access to the model dim by @joein

Fixes 🕸️

• #524 - fix propagation of specific_model_path and local_files_only to embed with parallel > 1 by @joein

v0.7.0

Changelog

Features 🏎️

• #513 - New sparse embeddings model with semantic understanding: MiniCOIL (Qdrant/minicoil-v1) by @generall

Fixes 🕸️

• #506 - fix list of supported languages in bm25 by @joein

v0.6.1

Changelog

Features 🏎️

• #490 - deprecate old archive struct when loading from custom urls in favour of model_name.tar.gz to ease adding custom models by @joein
• #492 - preserve embeddings in a type set by their model to allow lower precision outputs by @joein
• #496 - custom rerankers support by @I8dNLo

Fixes 🕸️

• #499 - fix splade hf source, causing model downloading errors in some cases by @joein

v0.6.0

Changelog

Features 🏎️

• #427 - introduce LateInteractionMultimodalEmbedding with ColPali-v1.3 by @I8dNLo @joein
• #463 - introduce dense text custom models (use a model which is not supported by fastembed if it follows the same preprocessing pipelines) by @joein
• #428 - add jina embedding v3 model by @hh-space-invader
• #443 - load a model from specific_model_path bypassing hf file structure by @I8dNLo
• #450 - extend type coverage, mark fastembed as typed package (also #451, #453, #454, #457, #458, #459, #460, #461, #464, #469, #470, #472) by @hh-space-invader @joein
• #440 - hide progress bar when a model is in cache by @hh-space-invader
• #455 - update to pillow<12.0.0 by @I8dNLo
• #484 - update to mmh3<6.0.0 by @joein

Fixes 🕸️

• #486 - fix thenlper/gte-large, use mean pooling of token embeddings instead of CLS embedding by @joein
• #436 - fix paraphrase-multilingual-MiniLM-L12-v2, do not normalize and use mean pooling of token embeddings instead of CLS embedding by @hh-space-invader
• #445 - fix paraphrase-multilingual-mpnet-base-v2 and intfloat/multilingual-e5-large, do not normalize and use mean pooling of token embeddings instead of CLS embedding by @I8dNLo

Thanks everyone who contributed to the current release!

v0.5.1

Change Log

Fixes 🐛

• #439 - fix python3.13 installation by making onnx an optional dependency, keep onnxruntime mandatory by @joein

v0.5.0

Features 📖

#403 - Drop Python 3.8 support by @joein
#404 - Add Python 3.13 support by @joein
#406 - Improve models cache progress bar by @hh-space-invader
#422 - Add multi-GPU example by @hh-space-invader
#425 - Provide user warning when specifying providers and CUDA by @hh-space-invader

---

Models 🧠

#405 - Support jina embeddings v2 models by @hh-space-invader
#408 - Add jina clip v1 model by @hh-space-invader
#415 - Added support for the thenlper/gte-base model by @hh-space-invader
#419 - Introduced parallel processing and pair-wise API for cross-encoders by @I8dNLo
#429 - All models now support Hugging Face (hf) compatibility by @I8dNLo

---

Fixes 🐛

#413 - Fix ColBERT model shape mismatch by @hh-space-invader

v0.4.2

Change Log

Features 📖

• #380 - add NOTICE file to support models with restrictive licenses by @hh-space-invader @joein
• #362 - unlock numpy v2 by @amietn
• #375 - add jina rerankers by @hh-space-invader

Models 🧠

• #378 - jinaai/jina-colbert-v2 by @hh-space-invader
• #375 - jinaai/jina-reranker-v1-tiny-en, jinaai/jina-reranker-v1-turbo-en, jinaai/jina-reranker-v2-base-multilingual by @hh-space-invader

Fixes 🐛

• #337 - temporarily restrict onnxruntime to <1.20 due to recent problems by @hh-space-invader

v0.4.1

Change Log

Features 📢

• #366 - replace pystemmer with py-rust-stemmers by @I8dNLo