Hello, I am using the newest ml-agents version(0.21.0) together with the unity obstacle tower environment (v 4.1). Previously I used an older version of ml agents (0.6.1) and the OTC environment (2.2) and automatically received visual observations of size 168x168x3. After upgrading the observations are now 84x84x3 and I can't figure out how to get the original size. When I set the --width and --hight argument when calling mlagents-learn it still resizes the observations to 84x84. Is there some other way I can set this parameter? My current environment setup: ml-agents: 0.21.0.dev0, ml-agents-envs: 0.21.0.dev0, Communicator API: 1.1.0, TensorFlow: 2.0.0 Thank you for your help!
UPDATE: I found out the problem was with the OTC environment executable I was using (https://github.com/Unity-Technologies/obstacle-tower-env/issues/126)