Using Llama2 to code generate opcodes

Hi Folks:

I am new to HuggingFace and Llama2. I want to use Llama2 code to generate “opcodes” for a simple bytecode interpreter I am working on. The idea would be I could give the model, JSON or natural language, and it would generate an “opcode.” I have rather small computer (16G, I5, 6 core, and NVIDIA 525.125.06). What would be the best way to start? Am I using some instruct model? Or fine tuning? Any advice would be helpful.