0.6.0

0.6.0

⭐ Highlight

rerankers goes multi-modal! We've overhauled the Document class to welcome a new family of rerankers, MonoVLM rerankers, with their first entry, MonoQwen2-VL-v.01, in #45.

📰 Other Changes

• Support for tokenizer_kwargs and model_kwargs, thanks to #44 by @sam-bercovici (also implemented as processor_kwargs for MonoQwen's image processor, following the same design pattern as the one by @sam-bercovici)
• No more prints on import and greater respect for passed verbosity (further changes at some point will improve over-verbosity.)

🛠️ Fixes

• Compatibility fixes for T5: the API for the transformers implementation that MonoT5 relies one has deprecated an argument. Our code now supports
• Better T5 test in #38 thanks to @eltociear
• Proper ordering of FlashRank ranker results in #46

0.5.2post1

ColBERT bug fixes & better support for answerdotai/answerai-colbert-small-v1

0.5.1

Minor Changes

• RankedResults is now subscribable

0.5.0

Changes

• Added support for the current state-of-the-art rerankers, BAAI's series of BGE layerwise LLM rerankers, based on Gemma and MiniCPM. These are different from RankGPT, as they're not listwise: the models are repurposed as "cross-encoders", and do output logit scores.

0.4.0

Release bundling up both the 0.3.1 minor release and the 0.4.0 release.

Changes

• ColBERT performance improvement! It should now be faster and result in stronger results following implementation of the JaColBERTv2.5 dynamic query length method.
• HuggingFace's Text-Embedding-Server (TEI) inference introduced as an API reranker option, thanks to @srisudarsan
• T5 bugfix for certain models, where the labels were incorrectly set. Thanks to @marcospiau
• Native default support for new Portuguese T5 rerankers introduced by @marcospiau

0.3.0

0.3.0 is here! It brings a lot of oft-requested features:

• A new transparent Document has been added. You may construct Documents yourself, or keep using the library exactly as-is. This object now allows for metadata support! You can pass a list of metadata (or add them to your Document objects) to rank() calls, and get the metadata back. Thanks to @Anmol6 for starting the work on this!
• RankLLM is now supported 🥳 RankZephyr and RankVicuna are implemented, but untested at the moment, while RankLLM + GPT models are fully supported. In version 0.5.0, this will become the default way of using GPT models for reranking purposes.
• Some QoL improvements, the most notable of which is that it is now possible to iterate directly on RankedResults objects rather than having to use the wordy for result in results.results.

0.2.0

• Introduce FlashRank support as suggested in #8
• Basic async methods by @tarunamasa #6
• Support the newly introduced MixedBread ranking API

0.1.1

Fixes issues with the langchain integration.

0.0.1post1

Initial release of rerankers, with the post1 minor dependency fix! (removing colbert-ai from [all] installs)

Features:

• Unified RankedResults and Result output format for all reranker types
• T5-based Seq2Seq pointwise rerankers (both MonoT5 and Inranker families)
• All sentencetransformers-friendly cross-encoders
• RankGPT
• ColBERT
• API-based rankers (Cohere, Jina)