I saw there is an imcomplete RNN implementation in the official sample. But it only contains one forward pass and one backward pass. It does not train on any real dataset. Also, it does not tell how to update the weight after the backward pass.
Is there any complete RNN/LSTM sample with cuDNN that trains on a real dataset? I want to do some optimization work on LSTM workload.I hope there is an RNN code that can be validated.