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 https://github.com/AnswerDotAI/rerankers/pull/45.

📰 Other Changes

• Support for tokenizer_kwargs and model_kwargs, thanks to https://github.com/AnswerDotAI/rerankers/pull/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 https://github.com/AnswerDotAI/rerankers/pull/38 thanks to @eltociear
• Proper ordering of FlashRank ranker results in https://github.com/AnswerDotAI/rerankers/pull/46