numpy dataset instead of tensor slices

Question

I am fitting my Tensorflow RNN+LSTM model with the time series data. The dataset is divided in 60:40 to get training and testing dataset respectively. Now, currently I am using