Intent-based Network Configuration with LLMs

• Prepair 2 ollama servers for running embedding and main LLM

  CUDA_VISIBLE_DEVICES=0 OLLAMA_HOST=0.0.0.0:11434 ollama serve
  CUDA_VISIBLE_DEVICES=1 OLLAMA_HOST=0.0.0.0:11435 ollama serve

• Create a secret.py file for openAI API key. Can put a dummy key if do not use chatGPT.

  echo 'OPENAI_API_KEY = "key"' > secret.py
  

• Run evaluation

  python main.py
  

Fine-tuning

• We use unsloth which support fine-tuning and output to ollama.

  python finetuning.py
  

How to add new use case

• Make a copy of example_use_case then modify the files to fit your use case.
• Modify main.py to add new use case.

Publication

To understand how this system work, please checkout our publication:

• Van Tu, Nguyen; Yoo, Jae-Hyoung; Hong, James Won-Ki; Intent-based Network Configuration using Large Language Models; International Journal of Network Management (IJNM, SCIE), vol. 35, Jan 2025, e2313.