---
base_model: AicoresSecurity/Cybernet-Sec-3B-R1-V0
tags:
- text-generation-inference
- transformers
- unsloth
- llama
- trl
license: llama3.2
language:
- en
---

## Model Description
AI Cores LLM that is fine tune with cybersecurity dataset for cybersecurity applications. It provides safety advise for python code, especially for Gen AI developers.

## Intended Use
- **Intended users**: Application Engineers, Software Engineers, Data scientists and developers working on cybersecurity applications.
- **Out-of-scope use cases**: This model should not be used for medical advice, legal decisions, or any life-critical systems.

# How to use


Use with transformers
Starting with transformers >= 4.43.0 onward, you can run conversational inference using the Transformers pipeline abstraction or by leveraging the Auto classes with the generate() function.

Make sure to update your transformers and bitsandbytes installation via pip install --upgrade transformers. pip install --upgrade bitsandbytes accelerate torch

Inference:

```python

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

# Create a BitsAndBytesConfig for 8-bit quantization
bnb_config = BitsAndBytesConfig(
    load_in_8bit=True
)

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AicoresSecurity/Cybernet-Sec-3B-R1-V0.1")

if torch.cuda.is_available():
    from transformers import BitsAndBytesConfig
    bnb_config = BitsAndBytesConfig(load_in_8bit=True)
    model = AutoModelForCausalLM.from_pretrained(
        "AicoresSecurity/Cybernet-Sec-3B-R1-V0.1",
        quantization_config=bnb_config,
        device_map="auto"
    )
else:
    # Fallback for CPU-only systems
    model = AutoModelForCausalLM.from_pretrained("AicoresSecurity/Cybernet-Sec-3B-R1-V0.1")


# Define your system prompt and user prompt
system_prompt = """You will be provided with a piece of Python code, and your task is to provide ideas for efficiency improvements ans safety. Only code
<reasoning>
...
</reasoning>"""

user_prompt = """
import os

def process_filename(filename):
    # Sanitize the filename to prevent path traversal attacks
    allowed_chars = "abcdefghijklmnopqrstuvwxyz0123456789_-"
    sanitized_filename = ''.join(c for c in filename if c in allowed_chars)
    
    if sanitized_filename != filename:
        raise ValueError("Invalid filename: only alphanumeric characters, underscores, and hyphens are allowed.")
    
    filepath = os.path.join("/user_data", sanitized_filename)
    
    try:
        with open(filepath, 'r') as file:
            content = file.read()
            return content
    except FileNotFoundError:
        return "File not found."

# Example usage
filename = "user_report.txt"
print(process_filename(filename))

filename_attack = "../../../etc/passwd"
try:
  print(process_filename(filename_attack))
except ValueError as e:
  print(e)
"""

full_prompt = system_prompt + user_prompt

# Tokenize the full prompt and move it to the model's device
input_ids = tokenizer.encode(full_prompt, return_tensors="pt").to(model.device)

# Generate output from the model
output_ids = model.generate(
    input_ids,
    max_length=100,         # Adjust as needed
    do_sample=True,         # Use sampling for more varied output
    temperature=0.7,        # Adjust for creativity
)

# Decode the generated tokens back into a string
output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print(output_text)
```

# Uploaded  model

- **Developed by:** AicoresSecurity
- **License:** apache-2.0
- **Finetuned from model :** AicoresSecurity/Cybernet-Sec-3B-R1-V0

This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.

[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)