LLM Client

This is a client for running experiments. Remember to update this readme when more abilities are added.

Experiments

Put all of your experiments in the experiments directy off root.

They're written in .yaml, and are self-explanatory. Or if they aren't self-explanatory just ask me: @jcronq.

Run Instructions

copy .env.example to .env Add your OpenAI API Key to .env install dependencies (probably all there, install using poetry)

poetry init
poetry install

Then to run an experiment

python main.py experiments/<your-experiment>.yaml

Results will be stored in <your-experiment>-results.json and <your-experiment>-results.csv. The .csv file might be broken after adding labels. (I like to use commas for lists!)

Operating cost

I run quite a few experiments. Costs < $0.02 / day on gpt-3.5-turbo. Not sure what gpt-4 will cost.