This model from McGill NLP applies the LLM2Vec method to convert Meta's Llama 3 8B Instruct model into a powerful text embedding model using modified next-token prediction (MNTP). It produces high-quality dense vectors for semantic search, clustering, and retrieval tasks. The resulting model is fully open and available on Hugging Face.
In the Foundation models & chat space, LLM2Vec Meta Llama 3 8B Instruct Mntp takes a focused approach. It focuses on creating strong unsupervised text embeddings from decoder-only LLMs without massive contrastive training. It is built as an open-source project for NLP researchers and embedding model users. The project is open source (MIT). LLM2Vec Meta Llama 3 8B Instruct Mntp is available on the web, the command line, and API.
Behind LLM2Vec Meta Llama 3 8B Instruct Mntp is McGill NLP, based in Canada, and it first shipped in 2024. The project is developed in the open on GitHub with 1.7k stars. Among its 4 catalogued features are Text Embeddings, Modified Next Token Prediction, and llama 3 Base.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do