Hello, I am working on a code that was written in TF1. I am studying the code and try to convert it into TF2. Today I struggle to know what is the equivalent of zero_state method from tf.compat.v1.nn.rnn_cell.RNNCell. I would like to be able to compute this zero-filled state tensor, but with native TF2 code (using keras).
Because, for me it is not clear how to make sure it is a tensor full of zeros in the RNN documentation.
Anyway I checked and this give same results as you. Also thanks for the link about TF1 to TF2 migration. I am a novice in TensorFlow, so beginning by trying to translate TF1 to TF2 is not the optimal way of learning it ! But I have to do it anyway, at least it is a way forcing me to go deep in the documentation.