Clear documentation on config dicts?

How severe does this issue affect your experience of using Ray?

  • High: It blocks me to complete my task.

I am trying to train a custom environment with a custom model both of which operate on graph neural networks.
I have debugged my environment to last iota, and I still receive the same error during check_env: the two structures are not the same.
Two data structures with the shape of my observation space are printed in the terminal, in which one is filled with rubbish (I have changed my returned observation to a fixed set of numbers and it checks ) and the other is empty, or rather, filled with dots.

That is why I believe now that the problem relies on the configuration I give to PPO.
Every example and piece of documentation I have seen gives a different implementation of the configuration, be it PPO_trainer, o, or whatever.

Any link or help in understanding how to get into using this library, in particular in debugging that weird error or simply learning the proper syntax or the difference between the difference flavours of syntaxes?

Please help me in taking the first steps. This is very confusing.

Hi @Maldades ,

Please provide your error at minimum.
Better would be a minimal repro script.