How to use a custom critic and default actor in PPO?

RLlib
1
  • High: It blocks me to complete my task.
2

@Matthew_Giacovelli thanks for posting and welcome to the forum.

For a custom value network you will need to write your own RLModule. See here for an example how this is done for Atari.

You can then build this RLModule via the PPOConfig by defining an RLModuleSpec similar to the same example here.