These datasets are used to evaluate models on French performance using: https://github.com/EleutherAI/lm-evaluation-harness