Prodigy SM Instruct v0.1-draft Overview
Prodigy SM Instruct v0.1-draft is a 7 billion parameter instruction-tuned model developed by draganjovanovich, building upon the Prodigy SM Base v0.1. Its primary focus is on robust instruction following, function calling, and tool usage across multiple languages, specifically Serbian, Croatian, Bosnian, and English. The model was instruction-tuned using a curated mix of high-quality datasets formatted in ChatML.
Key Capabilities
- Multilingual Instruction Following: Strong performance and precise instruction adherence in Serbian, Croatian, Bosnian, and English.
- Advanced System Prompt Handling: Capable of following extensive system prompts, including those over 500 tokens, particularly in Serbian.
- Function Calling & Tool Usage: Enhanced abilities for integrating with external tools and executing function calls.
- Conversational AI: Utilizes the ChatML format for effective conversational interactions.
Ideal Use Cases
- Applications requiring precise instruction following in Serbian, Croatian, Bosnian, or English.
- Multilingual conversational AI systems.
- Systems that leverage function calling and tool integration.
- General text generation in the supported languages.
Limitations
Like other large language models, Prodigy SM Instruct v0.1-draft may exhibit hallucinations or factual inaccuracies, and its performance in languages outside its primary focus may vary. Outputs should always be verified for critical applications.