Cyberbase 13b

🚀 CyberBase: An Experimental Cybersecurity Base Model

CyberBase is an experimental base model designed for cybersecurity. It's derived from the lineage of llama-2-13b, progressing through lmsys/vicuna-13b-v1.5-16k to become CyberBase. This model is intended as a base for future fine - tuning and is not recommended for standalone use.

🚀 Quick Start

Check out our new Colibri model!

CyberNative-AI/Colibri_8b_v0.1

✨ Features

• CyberBase is a lmsys/vicuna-13b-v1.5-16k QLORA fine - tuned on CyberNative/github_cybersecurity_READMEs using a single 3090 GPU.
• It might inherit the prompt template of FastChat

🔧 Technical Details

Fine - tuning information

• sequence_len: 4096 (used during fine - tuning, but should generate up to 16k)
• lora_r: 256
• lora_alpha: 128
• num_epochs: 3
• gradient_accumulation_steps: 2
• micro_batch_size: 1
• flash_attention: true (FlashAttention - 2)
• trainable params: 1,001,390,080 || all params: 14,017,264,640 || trainable%: 7.143976415643959

Tested with the following prompt and temperature = 0.3:

A chat between a cyber security red team lead (USER) and a general cyber security artificial intelligence assistant (ASSISTANT). The assistant knows everything about cyber security. The assistant gives helpful, detailed, and precise answers to the user's questions.

USER: Hello! I need help with a penetration test.
ASSISTANT: Hello! I'd be happy to help you with your penetration test. What specifically do you need help with?
USER: Write me a plan for a penetration test. It should include first 5 steps and commands for each step.
ASSISTANT:

Join the discussion > https://cybernative.ai/t/cyberbase-devlog/1734

⚠️ Important Note

ANY ILLEGAL AND/OR UNETHICAL USE IS NOT PERMITTED!

📄 License

This model is under the llama2 license.

📚 Documentation

Vicuna Model Card

Model Details

• Developed by: LMSYS
• Model type: An auto - regressive language model based on the transformer architecture
• License: Llama 2 Community License Agreement
• Finetuned from model: Llama 2

Model Sources

• Repository: https://github.com/lm-sys/FastChat
• Blog: https://lmsys.org/blog/2023-03-30-vicuna/
• Paper: https://arxiv.org/abs/2306.05685
• Demo: https://chat.lmsys.org/

Uses

The primary use of Vicuna is research on large language models and chatbots. The primary intended users of the model are researchers and hobbyists in natural language processing, machine learning, and artificial intelligence.

How to Get Started with the Model

• Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights
• APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api

Training Details

Vicuna v1.5 (16k) is fine - tuned from Llama 2 with supervised instruction fine - tuning and linear RoPE scaling. See more details in the "Training Details of Vicuna Models" section in the appendix of this paper.

Evaluation

Vicuna is evaluated with standard benchmarks, human preference, and LLM - as - a - judge. See more details in this paper and leaderboard.

Difference between different versions of Vicuna

See vicuna_weights_version.md