Code Llama Paper: https://arxiv.org/abs/2308.12950
Request License: https://ai.meta.com/resources/models-and-libraries/llama-downloads/
- Create a site specific file for your cuda environments. See existing site configurations.
pip3 install -r requirements.txt
. This will installjobrunner
./configure -s <site-name>
- Make sure
jobrunner
is in your path. jobrunner setup software/codellama -V
jobrunner setup models/codellama -V
. Follow instructions to provide license url and download desired models (7b,7b-Instruct,etc...)- Edit
Jobfile
in project root directory to set appropriate schedular commands. jobrunner submit runs/Example/completions -V
jobrunner submit runs/Example/instructions -V
- huggingface-cli login --token