Closed DavideSignori closed 1 year ago
It's very tedious, but you could modify some of the methods in demo_loader.
I created some utilities for modifying the demo files in my environment. You would need to make a few changes in order to get it to work for your environment though. (model name, etc.)
@miguelalonsojr I would also like mlagents to provide an easier way to modify the demo files after they are created. For example, suppose I spend 2 hours recording a demo for my agent, train, and then realize that one of the observations is causing problems. I don't want to have to re-record the demos again. It would be easier to just loop through all of the observations in the demo file and remove the problematic observation. I think that ml-agents should provide utilities that would make this possible.
Is your feature request related to a problem? Please describe. Hi I am encountering an issue with the demonstration recorder, i would like to be able to remove unnecessary information (actions or observations) from the .demo file after the recording took place. I have to do this because i can filter out unnecessary information only after the demonstration is completed. Is there a way i can do this?
Describe alternatives you've considered It would be even helpful if there was a way to perform such processing before the file is saved with the .demo extension. Thanks!