peft-cysec-qwen2.5

Model Card for peft-cysec-qwen2.5

This model is a fine-tuned version of Qwen/Qwen2.5-7B-Instruct on the CySecBench dataset. It is designed to classify cybersecurity-related prompts into one of ten attack categories.

Model Description

• Developed by: Erfan Sadeghiii
• Model type: A 7B parameter causal language model fine-tuned using QLoRA.
• Language(s) (NLP): English
• Finetuned from model: Qwen/Qwen2.5-7B-Instruct

Intended Uses & Limitations

This model is intended for research and educational purposes in the field of cybersecurity. It is designed to categorize prompts, not to generate harmful content.

How to Use

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load base model and the adapter
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-7B-Instruct", device_map="auto")
model = PeftModel.from_pretrained(model, "ErfanSadeghiii/peft-cysec-qwen2.5")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B-Instruct")

def predict(prompt):
    formatted_prompt = f"<start_of_turn>user\n{prompt}<end_of_turn>\n<start_of_turn>model\n"
    inputs = tokenizer(formatted_prompt, return_tensors="pt").to("cuda")
    with torch.no_grad():
        outputs = model.generate(**inputs, max_new_tokens=20, do_sample=False, temperature=0.0)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    category = response.split("<start_of_turn>model")[-1].strip()
    return category.split("<end_of_turn>")[0].strip()