How to get Belief Fine-Tuning working with RL Search?

[ravihammond]

Hi Hengyuan,

I'm wanting to get my agent to apply rl_search and belief fine-tuning at test time. I've successfully got rl_search working by running this python file, however, when I try to get rl_search to use a belief model using this argument, it's giving me an error on this line that the constructor for ARBeliefModel does not accept the "mode" keyword argument. It seems to me that the OBL github repository is not up-to-date to allow belief fine-tuning with RL search.

How can I get belief fine-tuning to work with rl search? Any help is greatly appreciated!

Thanks in advance :)

[hengyuan-hu]

Are you trying to run single-agent RL search or multi-agent RL search with belief fine-tuning? I don't think the multi-agent version is in the repo (unfortunately)... For the single-agent RL search, you can probably set the mode=private and see if the rest of the code works. We lost some code for the belief fine-tuning code so I haven't really tested it.