The RFT data

[ZIKEYUAN]

Hi,after completing SFT and multipath reasoning, I have some doubts about the data under the data/rft path in your github code base. I would like to ask you how these data are generated from? I see that four data sets are generated after the Filter reasoning path process, and I would like to ask whether the data under data/rft are created from four datasets?

[GanjinZero]

data/rft contains llama7b/13b/7b2/13b2 which means this dataset is generated by inferencing this size of SFT models with 100 times and filtered with correct and distinct reasoning paths.

[ZIKEYUAN]

Thank you, but I have a question. After the Filter reasoning path process, it will generate four files. If I don’t want to use the RFT data you provided, how can I use the four data files to generate my RFT data?

[GanjinZero]

I think one of the generated data is rft data which you can use directly. If you don't know use which one, you can copy some lines here and I will tell you.

[HaoRenkk123]

小白发问，我看推理生成的脚本中single_infer...py中的参数return_seq_num=1，这里是推理只生成一个样本对吗，为什么你说采样了100次呢，我有点没弄懂，我看执行完该脚本后，生成的raw_generation_{temp}_{seed}.json中也是只生成了一个推理结果