Output activations of individual neurons of a given layer of a trained model.

crazydonkey200 / tensorflow-char-rnn

Char-RNN implemented using TensorFlow.

MIT License

425 stars 267 forks source link

Output activations of individual neurons of a given layer of a trained model. #8

Closed tlhumphrey2 closed 6 years ago

tlhumphrey2 commented 7 years ago

I want to see what the activations are for individual neurons of a given layer for a given input character.

Any suggestions?

crazydonkey200 commented 7 years ago

The problem is that the current TensorFlow API for RNN doesn't expose the hidden activations. You probably need to create a customized RNN cell to add hidden activations into the output.

crazydonkey200 commented 6 years ago

Close for now. You could reopen it when there are further questions.