sweetice/Deep-reinforcement-learning-with-pytorch

I dont think PPO pendulum is converging

Bigpig4396 opened this issue · 4 comments

Bigpig4396 commented
I dont think PPO pendulum is converging
KT27-A commented

Yes, the problem is that the activation function is chosen incorrectly.

  • 👍1
HuangHaoyu1997 commented

I don't think this repo implement the PPO correctly either

NanJuni commented

change the activation function relu to tanh

wiluen commented

right,change relu to tanh in actor network