farizrahman4u / recurrentshop

Framework for building complex recurrent neural networks with Keras
MIT License
765 stars 218 forks source link

Error when combining teacherforce/readout with return_states #107

Open Aerylia opened 6 years ago

Aerylia commented 6 years ago

I get an error stating that the output_tensors and output_masks are not of equal length when using readout and return_states and nesting the RecurrentModel into a keras model. I believe this is because RecurrentModel.compute_masks(..) does not take into account that the number of states is 1 more than the number of outputs when self.readout = True. My error went away when I fixed this, but I am not sure how this will affect the remainder of the code/models/layers etc.