MoMo

About

The MoMo family of large language models, developed by Kibong Choi, features a range of models built upon the MoMo 70B v1.1 foundation. These models are designed for flexibility and are available in various quantized formats such as GGUF, GPTQ, and AWQ, each offering a balance among size, speed, and performance. The GGUF versions, for instance, support quantization levels from 2-bit to 8-bit, enabling users to select the model configuration most suited for their computing capabilities. They integrate seamlessly with multiple inference clients and libraries, including tools like llama.cpp, text-generation-webui, and ctransformers. Originally, the MoMo 70B v1.1 was trained using Orca-style and Alpaca-style datasets, but its quantized iterations provided by TheBloke on Hugging Face make it more accessible to a diverse user base. It's important to differentiate this MoMo family from the Molmo family, which is a separate initiative by the Allen Institute for AI.