Loading the reacher-easy task I received an error about the observation dimensionality (6 vs. 7).
The original DeepMind paper states that the observation dimensionality is 7, which is also set in MBBL's env_register.py. However, debugging into dm_control/suite/reacher.py, the observations seem to have 6 dimensions (2 pos, 2 vel, 2 goal).
I was wondering whether this is an issue on my end or whether running the reacher-easy task also fails for you? Thanks a lot!