Justin Chu

Cc @gramalingam

#5019 in progress. We still need (3) from your suggestions and the ability to register the op. I may not be able to pick them up soon. Contributions welcome and...

Please feel free to review https://github.com/onnx/onnx/pull/5906

@p-wysocki

Thanks! You may use https://onnx.ai/onnx/operators/text_diff_Split_13_18.html as a reference

Is it possible when the second input to split is `""`, which will give `node->inputs().size() == 2` and skip the logic? In ONNX, a trailing optional input (in this case...

@ZelboK The script I used ```python import onnxscript from onnxscript import opset11 as op from onnxscript import FLOAT import onnx from typing import Tuple @onnxscript.script() def model(input: FLOAT[12, 42]) ->...

From here ``` < ir_version: 6, opset_import: ["" : 18] > model (float[12,42] input) => (float[6,42] a, float[6,42] b) { a, b = Split (input) } ``` We see the...

Model checker does not raise an error even when we have this line https://github.com/onnx/onnx/blob/6eb45c0e08b819deacac47c4099945f13e784a3a/onnx/defs/tensor/defs.cc#L734 I wonder what happened?

@fatcat-z