somosnlp/gemma-FULL-RAC-Colombia

Gemma-FULL-RAC-Colombia: Specialized for Colombian Aviation Regulations

Gemma-FULL-RAC-Colombia is a 2.6 billion parameter causal language model, developed by Edison Bejarano and Nicolai Potes, and funded by Fundación Universitaria Los Libertadores. It is fine-tuned from Google's Gemma model using LoRA (Low-Rank Adaptation) to specialize in the Colombian Aeronautical Regulations (RAC) dataset. This adaptation ensures a deep understanding of Colombian aviation terminology, concepts, and regulatory frameworks, while retaining Gemma's broad language capabilities.

Key Capabilities

• Domain-Specific Understanding: Profound grasp of Colombian aviation law and regulations.
• Text Generation: Capable of generating text and insights related to the RAC.
• Question Answering: Provides answers to queries concerning Colombian aeronautical regulations.
• Language: Tailored for the Colombian legal and regulatory context in Spanish.
• Efficient Fine-tuning: Utilizes LoRA for efficient adaptation, preserving base model knowledge.

Good For

• Direct Use: Generating text, answering questions, and providing insights on Colombian Aeronautical Regulations without further fine-tuning.
• Educational Content: Creating materials to simplify and explain aviation legal language.
• Regulatory Compliance: Assisting in efforts to ensure adherence to Colombian aviation laws.
• Downstream Applications: Integration into systems for automated compliance checks, document summarization, and training simulators (with further fine-tuning).

Limitations

• May not cover nuances of aviation regulations outside of Colombia.
• Requires updates to adapt to changes in law.
• Not designed for navigational purposes or to replace professional legal advice.