I noticed there was a previous effort to add support for ColBERT-style multi-vector embeddings (#209).
Would it be possible to support the following Hugging Face models?
- lightonai/LateOn
- LiquidAI/LFM2.5-ColBERT-350M
These are both performant late-interaction embedding models remaining practical to run locally. It would be great to be able to use them with fastembed for local semantic search and RAG applications.
Thanks.
I noticed there was a previous effort to add support for ColBERT-style multi-vector embeddings (#209).
Would it be possible to support the following Hugging Face models?
These are both performant late-interaction embedding models remaining practical to run locally. It would be great to be able to use them with fastembed for local semantic search and RAG applications.
Thanks.