Should batch size matter while implementing LSTM inference in hardware?

Question

I\'m trying to implement a recurrent neural network model (LSTM) on hardware, the model was trained in Pytorch to get the required results. I\'m trying to abstract the high-leve