There is a possibility to add extra info in the configuration dictionary or Rllib. For example to change the folder name in which ray saves the results.
I’m testing different implementations of my environment and I would like to save some kind of comment that allows me to differentiate the various modifications.
Is there any way to achieve that already implemented in ray?
Hi @tadteo ,
and welcome to the discussion board. It is a little unclear what you exactly want to do. Do you want to add to the
info dict returned by the envrionments
step() function? Or do you want to the add keys to the configuration dictionary defined for each
Sorry, I was not clear
What I was trying to do was to modify the config dictionary to add “info” on the specific training.
In particular, I was trying to save a “map_name” parameter of the environment.
So I was wondering if it is feasible to save this name for example in the params.json file that is inside ~/ray_results/<experiment_name> or directly writing that name at the end of the <experiment_name> folder.
However, also other approaches to saving this kind of information are good.