/human-eval

Primary LanguagePythonMIT LicenseMIT

HumanEval for GPT-3.5/GPT-4

Results are here. Forked from OpenAI's repo.

To generate the completions (after pip installing the requirements), run:

mkdir results
python run.py

Then to evaluate the completion results, run

pip3 install -e .
python -m human_eval.evaluate_functional_correctness results/name_of_results_file