Open lumelanie opened 4 years ago
Hello, You need to use a custom version of stable-baselines for that.
Okay thanks, you mean defining my own custom network?
Bit more than that. Unless you find a way to obtain activations from layers without modifying the session.call
, you also have to modify those calls to include outputs for those mid-network activations. I have not touched TF1 in a while, and I am not sure if this is possible.
Hello :) Is there a way in Tensorflow to get the activations from the SAC actor network (as you could do by registering a hook to the layer in PyTorch)?