Overview

pki/nova-8b-cybersec is an 8 billion parameter language model, fine-tuned from the cognitivecomputations/Dolphin3.0-Llama3.1-8B base model. It has been specifically optimized for various cybersecurity tasks through QLoRA fine-tuning, utilizing a rank of 64 and an alpha of 128. The model was trained over 5 epochs with an effective batch size of 40, processing 40,075 training examples.

Key Capabilities

• Cybersecurity Expertise: Specialized in understanding and generating content related to cybersecurity, trained on datasets like SecurityGPT and PKI Context QA.
• Contextual Understanding: Supports a context length of 8192 tokens, enabling it to process substantial security-related documents and queries.
• Diverse Training Data: Fine-tuned on a comprehensive set of cybersecurity-focused datasets, including document summaries, Elbranschen threats, and ISO 27001/27005 controls.
• ChatML Format: Designed to work efficiently with ChatML formatting for conversational applications.

Good For

• Security Question Answering: Answering queries related to cybersecurity concepts, threats, and protocols.
• Document Summarization: Summarizing security reports, incident analyses, and compliance documents.
• Threat Intelligence Analysis: Assisting in the analysis of cybersecurity threats and vulnerabilities.
• Compliance and Standards: Understanding and interpreting information related to ISO 27001 and ISO 27005 standards.