DDPG doesn't work

[DennisWangCW]

I downloaded the baselines code, installed it using pip install -e ., and ran DDPG using the proposed command line. I've tested it on several mujoco environments but none of them converged at all. I didn't edit the code and want to know what's wrong with it? Hope for reply, thanks.

[DanielTakeshi]

It's helpful if you can provide more information and the log files as well. For example, see the issue I filed a while back about DQN:

#431

[QiXuanWang]

I didn't run Mujoko but tried some other popular envs (MountainCar for example), which should converge since it's so simple. But default training with 1e6 steps give me big negative returns. Not sure what to tune... While some other implementation gave me positive return though probably the calculation method is slightly different.