Matter-0.1-7B-boost Overview

The 0-hero/Matter-0.1-7B-boost is a 7 billion parameter language model, building upon the Mistral 7B architecture. It has undergone a full fine-tuning process using the proprietary Matter dataset, which is an extensive collection derived from over 35 distinct datasets, encompassing an analysis of more than 6 billion tokens. This "boost" version includes additional fine-tuning data to enhance its capabilities.

Key Capabilities

• Extensive Training Data: Fine-tuned on the Matter dataset, a large-scale, curated collection of over 6 billion tokens from diverse sources.
• Function Calling Support: Designed to integrate with external tools and APIs through explicit function calling mechanisms. It utilizes special tokens (<|begin_func|>, <|end_func|>, <|begin_func_response|>, <|end_func_response|>) to manage function calls and responses within the conversation flow.
• ChatML Format: Employs the ChatML prompt format, ensuring compatibility with common conversational AI frameworks.

Good For

• Conversational AI: General-purpose chat applications and assistant roles.
• Tool-Augmented Applications: Use cases requiring the model to interact with external functions, such as fetching real-time data, executing commands, or integrating with other software services.