d3rlpy
d3rlpy copied to clipboard
Can we add info to dataset?
I'm trying to modify the cql to adjust our case. In our scenaro, we have varying number of available actions per state so try to use action mask to mask out the invalid actions. In that case, I need the either store in observations as dictionary or store with info.
@AprilXiaoyanLiu Hello, thanks for the issue. Sorry, I don't understand your use case. What do you mean by mask out the invalid actions
? Is the action different from the dataset actions?
@takuseno In my use case scenario, different state has different available actions. So at each state, I hope to mask out the actions that are invalid/not available for each state.