Overview

M-SOLAR-10.7B-v1.3 is a 10.7 billion parameter language model developed collaboratively by megastudyedu, Prediction, and MICE. It is an iteration built upon the foundation of the megastudy/M-SOLAR-10.7B-v1.1-beta model.

Key Enhancements

• Expanded Dataset: The model has been further trained by incorporating additional, processed in-house datasets beyond those used for its base model.
• Data Augmentation: The quantity of data generated through random shuffling has been increased, contributing to a more robust training regimen.

Intended Use

This model is suitable for a variety of natural language processing tasks, leveraging its 10.7 billion parameters and refined training data to offer improved performance over its base version. Its enhancements suggest a focus on general-purpose language understanding and generation, making it a versatile tool for developers.