Is it possible to load checkpoints to evaluate the model?

[zengjia121]

Hello, thank you very much for the code you provided, I am very interested in it and I have a question when trying it.

1. I saw that there is eval_KGSimp in the file, but it does not seem to be used in the main program. Is this file used to evaluate the generated model?
2. I want to know if I can simply load the checkpoint and then run the model. evaluation, because from the main.py it seems that I have to set the do_train option
3. I have tried to write the relevant parameters and run the eval function, but there will be situations where "scoring.guardrails" and other packages cannot be imported, but I found it I'm wondering if I'm running this in the wrong place