Paschalidis-NOC-Lab/PandemIQ-Llama

PandemIQ-Llama: Domain-Adapted for Pandemic Intelligence

PandemIQ-Llama is an 8 billion parameter large language model (LLM) with a 32,768 token context length, developed by Paschalidis-NOC-Lab. It is a specialized variant of Llama-3.1-8B, continuously pre-trained on an extensive dataset of 5.8 billion tokens focused on pandemic-specific text.

Key Capabilities

• Domain Adaptation: Optimized for understanding and generating text related to pandemics, biothreats, and public health intelligence.
• Foundation for BEACON: Serves as the core AI component for the BEACON platform, an open-source initiative for biothreats surveillance and response.
• Llama-3.1 Base: Benefits from the robust architecture and general language understanding of the Llama-3.1 series.

Good For

• Pandemic Intelligence: Analyzing and synthesizing information from diverse sources related to disease outbreaks, public health, and biosecurity.
• Biothreats Surveillance: Supporting early detection and monitoring of emerging biothreats.
• Research and Development: Researchers and developers working on applications requiring specialized knowledge in epidemiology, virology, and public health communication.
• Customization: Easily fine-tunable using LoRA for specific downstream tasks within the pandemic intelligence domain.