code for generate data for policy and PRM?

[Xccanxin]

Hi, thanks for your great work. But i am confused about the data-collecting code.
As you mentioned:

1. Download policy data (positive samples) for training 1st policy model (Llama3-8b-Instruct): [Hugging Face]

2. Download PRM data (positive and negative samples) for training 1st reward model (Mistral-7B: MetaMATH): [Hugging Face]

how can i get these two data from your code? Is it from codes like self_train/generation/generate_both_samples_MATH.py or evaluate.py?
what is the key parameters to change to get these two data?

[zhangdan0602]

Thank you for your question! We have provided all links to policy datasets (see https://github.com/THUDM/ReST-MCTS#policy-data).

[Xccanxin]

Thank you for your question! We have provided all links to policy datasets (see https://github.com/THUDM/ReST-MCTS#policy-data).

thanks for the reply! how can i replicate these data myself by running the code rather than directly download？

[sxpro]

how can i replicate these data myself by running the code rather than directly download？

I have same questions. I wanner know how to obtain "zd21/ReST-MCTS-PRM-0th" along with code examples.

[Xingxiangrui]

+1 for this.

[RewindL]

I have same question.