Dataset & evaluation pipeline for hallucination detection in zh-tw
Python
0
24 commits
updated Jun 26, 2024
{modelname}.py,依所需更改輸入輸出檔名和 HuggingFace 資料 & 模型的 access tokenouts/{modelname}-res.pkl,格式為 [{第一篇回應}, {第二篇回應}, ...]python gpt4-judge.py {modelname} (7b, 8b or breeze)outs/eval-{modelname}.txt,格式如下[Label]
Yes/No
[Label]
Yes/No
...
=====
(下一篇的判斷)
outs/eval-{modelname}.txt) 重新算分python score.py {modelname} (7b, 8b or breeze)Dataset & evaluation pipeline for hallucination detection in zh-tw
Python
0
24 commits
updated Jun 26, 2024
{modelname}.py,依所需更改輸入輸出檔名和 HuggingFace 資料 & 模型的 access tokenouts/{modelname}-res.pkl,格式為 [{第一篇回應}, {第二篇回應}, ...]python gpt4-judge.py {modelname} (7b, 8b or breeze)outs/eval-{modelname}.txt,格式如下[Label]
Yes/No
[Label]
Yes/No
...
=====
(下一篇的判斷)
outs/eval-{modelname}.txt) 重新算分python score.py {modelname} (7b, 8b or breeze)