Laboratory Notebook for LLM-Based Code Engineering for Scientific Computing Applications

Code Llama Paper: https://arxiv.org/abs/2308.12950

Request License: https://ai.meta.com/resources/models-and-libraries/llama-downloads/

Setting up Code Llama

1. Create a site specific file for your cuda environments. See existing site configurations.
2. pip3 install -r requirements.txt. This will install jobrunner
3. ./configure -s <site-name>
4. Make sure jobrunner is in your path.
5. jobrunner setup software/codellama -V
6. jobrunner setup models/codellama -V. Follow instructions to provide license url and download desired models (7b,7b-Instruct,etc...)
7. Edit Jobfile in project root directory to set appropriate schedular commands.
8. jobrunner submit runs/Example/completions -V
9. jobrunner submit runs/Example/instructions -V

Hugging face

1. huggingface-cli login --token