This repository contains the code used in Will we run out of data? Limits of LLM scaling based on human-generated data
Directory v1 contains the code used in the October 2022 version of the paper.
First buid and start the container with
docker build -t data-stocks .
docker run -d -v ./results:/usr/src/app/results --name data-stocks-cont data-stocks
Then, execute any script you want with
docker exec your-container-name python src/script_name.py
The plots will be saved in the results directory. To generate all the plots in the main body of the paper, run stock_adjust.py.