PyTorch custom-op with None return type does not show up in torch-mlir

1

Hi all,
I’m trying to add a custom-op to PyTorch that looks as follows:

myLib.define("custom_op(Tensor t) -> None")

@impl(myLib, "custom_op", "CompositeExplicitAutograd")
def custom_op_impl(t: torch.Tensor) -> None:
    return

The above op never shows up when I lower my PyTorch module to torch-mlir. However, if I change the op signature to return a torch.tensor things things start working.

Is it not possible to create a custom op with a None return type?