I don’t understand if is possibile with VTA do a deploy of a Recurrent Neural Network (RNN), since all tutorial are only on CNN.
If you can unroll the RNNs, it should in principle be similar to the CNN tutorials. I don’t know if the current VTA code generation would be able to support sequential input (I’ve played around with VTA’s software simulator for my research but I haven’t used the code generation very much).