BatchToSpaceND: support 3d inputs #572
Description
I'm trying to convert the OpenSeq2Seq's wav2letter model checkpoints from TensorFlow to ONNX.
It contains BatchToSpaceND ops. Inputs are however a 3d tensor. Currently tfonnx supports NHWC.
Proposal: if the input is 3d, insert automatically an Unsqueeze op or a rewriter in order to reshape to NH1C and then back to NHC.