I am working with RLlib and ray tune.(training Deep reinforcement learning)
I want to check policy network function and computing process of outputing action, but I can’t find where.
Where can I check transfer process of training data and computing process of policy network?
Can you elaborate on what you mean by computing process?
RLlib only goes so far as to call functions of underlying tensor backend libraries.
These are closely tied to highly optimized code that is not easily accessible to view such progress.
In between defining inputs, making a call to, let’s say, Tensorflow and receiving the outputs from that operation, there is no way to view anything really.
The closest thing you can do is to define your model, run a training calling ray.init() with local_model=True and step into your model while executing eagerly, following the flow of tensors there.
Cheers
This also gives you a way of timing these calls if you are interested in that.