seqio advertises its compatibility with JAX and PyTorch in the README. However, the only docs I have found use tf.Tensors as inputs and outputs for Pre- and Postprocessors. How does one create a Preprocessor that operates on jax.Arrays, np.ndarrays, or anything else?
