Model Overview

maywell/Synatra-7B-v0.3-base is a 7 billion parameter language model built upon the mistralai/Mistral-7B-Instruct-v0.1 base. Developed by maywell, this model has undergone specific fine-tuning to enhance its capabilities in role-playing (RP) scenarios and general knowledge tasks. The training was conducted using 8 A6000 48GB GPUs, focusing on refining datasets and improving common sense understanding.

Key Capabilities

• Role-Playing (RP) Optimization: Explicitly tuned for role-playing applications.
• Enhanced Common Sense: Improvements in general knowledge and common sense reasoning.
• Instruction Following: Supports both ChatML and Alpaca (No-Input) instruction formats, making it versatile for various prompting styles.
• Non-Commercial Use: Licensed under CC BY-NC 4.0, strictly for non-commercial applications.

Good For

• Developers and researchers looking for a 7B model optimized for creative role-playing interactions.
• Applications requiring a model with improved general knowledge and common sense understanding.
• Experimentation with ChatML and Alpaca instruction formats in a non-commercial setting.