XRNN Runtime expects tensors with the shape = BxNxW, where:
- B is the batch size
- N is the number of frames
- W is the number of 16-bit words in each frame aligned to the tile array of the IP
runner->get_input_tensors() and runner->get_output_tensors().However, in RNN applications, the number of frames in each input can vary. So, only the B and W obtained from the tensor shape are useful to the user while N should be based on the actual input or output. You must create input and output tensors with aligned dimensions.