search

nlpxucan / WizardLM

LLMs build upon Evol Insturct: WizardLM, WizardCoder, WizardMath
9.19k stars 713 forks source link

evaluate_functional_correctness ${output_path}.jsonl 得到的pass@k结果全是1 #87

Closed WUHU-G closed 1 year ago

WUHU-G commented 1 year ago

image

ChiYeungLaw commented 1 year ago

I think you first need to check whether you have generated the codes correctly. Then check whether you install the humaneval environment correctly. Since your capture only provides limit information, it is hard for me to provide a solution.