How to parse a function with variadics ?

[Johansmm]

I would like to know if I can make inference in a model that contains multiple inputs in some functions

Example of the issue :

import onnx
import onnx.parser
import onnxruntime as ort

model = onnx.parser.parse_model(
    """
    <ir_version: 8, opset_import: [ "" : 15, "local" : 1 ]>
    agraph (float[N, M] A) => (float[?, M] X, float[?, M] Y, float[?, M] Z){
        X = local.concat(A, A)
        Y = local.concat(A, A, A)
        Z = local.concat(A, A, A, A)
    }

    <opset_import: [ "" : 15 ], domain: "local">
    concat (a, b, c, d) => (z) {
        z = Concat<axis=0>(a, b, c, d)
    }
    """
)
sess = ort.InferenceSession(model.SerializeToString())

feed = {"X": np.random.uniform(0, 1, size=(1, 2)).astype("float32")}
# Here, onnxruntime raise an exception: input != nullptr was false. input count mismatch
output = sess.run(None, feed)

How should I correctly write the local.concat function to validate a set of dynamic inputs?