TF_OperationDescription* desc = TF_NewOperation(graph, "Split", name);
  TF_AddInput(desc, {zero, 0});
  TF_AddInput(desc, {input, 0});
  TF_SetAttrInt(desc, "num_split", 3);
  TF_SetAttrType(desc, "T", TF_INT32);
  // Set device to CPU since there is no version of split for int32 on GPU
  // TODO(iga): Convert all these helpers and tests to use floats because

	// Create a split function that delivers a little data, then a predictable error.
	numSplits := 0
	const okCount = 7
	errorSplit := func(data []byte, atEOF bool) (advance int, token []byte, err error) {
		if atEOF {
			panic("didn't get enough data")
		}
		if numSplits >= okCount {
			return 0, nil, testError
		}
		numSplits++
		return 1, data[0:1], nil
	}
	// Read the data.

`tf.split(value, num_or_size_splits, axis)`. * `tf.sparse_split` now takes
arguments in reversed order and with different keywords. In particular we now
match NumPy order as `tf.sparse_split(sp_input, num_split, axis)`. NOTE: we have
temporarily made `tf.sparse_split` require keyword arguments. * `tf.concat` now
takes arguments in reversed order and with different keywords. In particular we