I was wondering if it would be possible to release the code that was used to generate the datasets for the gym and atari experiments. That would facilitate the evaluation of decision transformer methods on other environments.
This would be really good. Furthermore, regarding the Key-to-Door environment mentioned in the paper, I'd be interested to see the code for that environment and the algorithm used to generate trajectories on it.
I was wondering if it would be possible to release the code that was used to generate the datasets for the gym and atari experiments. That would facilitate the evaluation of decision transformer methods on other environments.