Keras tensorflow model convertion


I am trying to implement Keras model to TVM, but my approach is different which is. I create a model in Keras and convert it to TensorFlow and then try to import it to TVM. But here I use recurrent layers which converting (for ex: LSTM layer generates a tensorarray inbuilt conversion ). So can anyone tell me is this a valid way or it doesn’t work ?.

Also, I know LSTM works on KERAS FRONTEND way…

TensorFlow dynamic RNNs are not working yet. If you get them working, please send a PR :slight_smile:

Ok thanks, will try it out