How to get the results of Figure-3 (turns to capture) from the paper

[kailashg26]

Hi,

Thanks for the implementation and wonderful work. I was trying to run the IDDPG with a simple tag and trying to observe the result. Can you help me understand how the Fig-3 is plotted in the paper? Is it that you check the entropy, or is there any other term I'm missing here?

Also, I have another question: what should we do to observe the training progression on the terminal instead of saving it? Can you also provide a command to monitor the training process?

Thanks,

Kailash

[Xiyou521]

Hello, may I ask if you have converged the environment when running "simple_spread"? I am unable to converge when running this environment.

[hsvgbkhgbv]

The result should display zig-zag and the plot in the paper just shows the smoothed curve. Hope this can help you!

[Xiyou521]

I'm sorry to trouble you again. The problem I encountered was not smoothness, but rather the non convergence of rewards, and the values of "Action Loss, Value Loss, Entropy" also fluctuated greatly. I haven't made any changes to the parameters, I just modified the expression of success, which shouldn't involve any policy update issues. I really want to restore your code. Do you still want to save the original code data?