search

IanYangChina / DRL_Implementation

A DRL implementation repo
MIT License
19 stars 6 forks source link

Train Problem #2

Closed scw318 closed 3 years ago

scw318 commented 3 years ago

Thank you for your reinforcement learning algorithm and environment implementation. However,when I used the example about KukaPushPHER , the avg return was from -40 to -50, and the success rate from 0.0 to 0.2. I did not modify any parameters and I did’t know why this happened.Can you give me some advice about it?Thank you a lot.

IanYangChina commented 3 years ago

Thank you for your reinforcement learning algorithm and environment implementation. However,when I used the example about KukaPushPHER , the avg return was from -40 to -50, and the success rate from 0.0 to 0.2. I did not modify any parameters and I did’t know why this happened.Can you give me some advice about it?Thank you a lot.

Hi, thanks for the report. The PMG environment and the HER algorithm are sometimes sensitive to random seeds. You can try to run with more different random seeds and see how it goes. Check out this repo as well.

scw318 commented 3 years ago

Thank you very much, I will try it.