ExtremITA Camoscio 7B: Italian Instruction-Tuned LLaMA Model

ExtremITA Camoscio 7B is a 7 billion parameter language model developed by sag-uniroma2, building upon the LLaMA architecture. It is specifically fine-tuned for processing and generating content based on Italian instructions.

Key Characteristics

• Base Model: Derived from the teelinsan/camoscio-7b-llama adapters and the original LLaMA model.
• Italian Focus: Optimized for tasks requiring understanding and generation in the Italian language.
• Merged Adapters: This version integrates the adapters directly into the model, creating a more stable and robust base for subsequent fine-tuning efforts.
• EVALITA 2023 Participation: The model was further fine-tuned for the EVALITA 2023 challenge, indicating its suitability for competitive NLP tasks in Italian.

Use Cases

This model is particularly well-suited for:

• Italian NLP Applications: Any task requiring strong performance in Italian language understanding and generation.
• Instruction Following: Generating responses or completing tasks based on explicit instructions provided in Italian.
• Further Fine-tuning: Its stable, merged architecture makes it an excellent foundation for domain-specific or task-specific fine-tuning within the Italian language context.