Matter-0.1-Slim-7B-A Overview

Matter-0.1-Slim-7B-A is a 7 billion parameter language model, a fine-tuned variant of Mistral 7B, developed by 0-hero. It was trained for approximately 15 hours over 3 epochs on 4x A100 GPUs using Axolotl, leveraging the Matter-0.1-Slim-A dataset. This dataset comprises around 285,000 rows, meticulously curated from over 35 distinct datasets that collectively analyzed more than 6 billion tokens.

Key Capabilities

• Function Calling: The model natively supports function calling, utilizing special tokens (<|begin_func|>, <|end_func|>, <|begin_func_response|>, <|end_func_response|>) for structured interaction with external tools or APIs.
• ChatML Format: It adheres to the ChatML prompt format, ensuring compatibility with common conversational AI frameworks.
• Specialized Training Data: Fine-tuned on a highly curated dataset, suggesting potential optimization for specific domains or tasks represented within the Matter dataset.

Good For

• Applications requiring integrated function calling for tool use or API interaction.
• Developers seeking a Mistral 7B-based model with specialized fine-tuning.
• Use cases benefiting from a model trained on a diverse, large-scale curated dataset.