TextBase-7B-v0.1-GGUF

This is quantized version of SF-Foundation/TextBase-7B-v0.1 created using llama.cpp

Model Description

Finetuned version of Mistral-7B-Instruct. Details on development to be published soon.

TextBase-7B was fine-tuned from the open source Mistral-7B model using a novel and patent-pending learning technique. Our learning framework relies on efficiently combining supervised and Reinforcement learning methods leveraging human and AI labels over a combination of public and CRM task-specific datasets. Supervised finetuning allows the model to learn task-specific skills while RLHF imparts human judgement making the model able to generalize, reason and follow instructions efficiently.

Checkout more models developed by Salesforce under https://huggingface.co/Salesforce