Cannot reproduce 'model_exp' for toy arm environment

[bretthannigan]

Hello,

I have osim-rl running and am trying to get similar results to the provided examples/model_exp controller on the toy arm environment. As described here, I understand the command:
python -m examples.train_arm --train --steps 100000 --model examples/model_exp
should reproduce similar behaviour to the included example controller. When I run this command, I get a controller that does not seem to track the target at all and appears to produce similar movements regardless of the target location. The included examples/model_exp does much better. Is there anything I am missing?

Thank you!