The "split" passes result in invalid models.

Recently, I noticed that the split pass does not work, both in init and predict operations.
In particular, it results into invalid model parts. I would expect that each part would be a valid net, but this is not the case. I invoked them separately to determine which part plays a role to it, and both passes result in invalid nets.

For example (this is encountered in numerous models):

SSD (ONNX Hub, opset=12)

For split_init pass:
"split_init": "IndexError('list index (0) out of range')"

For split_predict pass:
"split_predict": "ValueError(\"Required inputs (['Constant_724', 'Constant_723', 'Constant_721', 'Constant_711', 'Constant_708', 'Constant_698', 'Constant_696', 'Constant_694', 'Constant_693', 'Constant_692', 'Constant_691', 'Constant_689', 'Constant_685', 'ConstantOfShape_680', 'Constant_675', 'Constant_672', 'Constant_662', 'Constant_648', 'Constant_647', 'Constant_644', 'Constant_643', 'Constant_642', 'Constant_722', 'Constant_641', 'Constant_638', 'Constant_636', 'Constant_631', 'Constant_630', 'Constant_623', 'Constant_622', 'Constant_684', 'Constant_621', 'Constant_617', 'Constant_613', 'Constant_609', 'Constant_604', 'Constant_603', 'Constant_602', 'Constant_601', 'Constant_598', 'Constant_597', 'Constant_593', 'Constant_701', 'Constant_591', 'Constant_588', 'Constant_627', 'Constant_585', 'Constant_582', 'Constant_578', 'Constant_577', 'Constant_576', 'Constant_574', 'Constant_573', 'Constant_571', 'Constant_569', 'Constant_568', 'Constant_608', 'Constant_566', 'Constant_562', 'Constant_561', 'Constant_560', 'Constant_556', 'Constant_551', 'Constant_539', 'Constant_538', 'Constant_536', 'Constant_534', 'Constant_639', 'Constant_531', 'Constant_528', 'Constant_525', 'Constant_522', 'Constant_520', 'Constant_518', 'Constant_612', 'backbone.model.layer2.0.3.bn2.running_var', 'backbone.model.layer1.4.2.bn2.running_mean', 'Constant_544', 'backbone.model.layer2.0.3.bn2.running_mean', 'backbone.model.layer1.0.weight', 'backbone.model.layer2.0.2.conv2.weight', 'backbone.model.layer2.0.2.bn1.weight', 'Constant_686', 'Constant_553', 'backbone.conf.5.weight', 'Constant_527', 'backbone.model.layer2.0.2.conv1.weight', 'Unsqueeze_412', 'backbone.model.layer2.0.4.conv1.weight', 'backbone.loc.5.weight', 'ConstantOfShape_678', 'backbone.model.layer2.0.1.bn2.running_mean', 'Constant_501', 'Constant_649', 'backbone.model.layer2.0.1.bn1.running_var', 'backbone.model.layer1.4.2.conv2.weight', 'backbone.model.layer1.5.2.bn2.running_var', 'Constant_614', 'backbone.model.layer2.0.1.bn1.weight', 'backbone.model.layer2.0.1.conv1.weight', 'backbone.model.layer1.5.0.bn2.bias', 'backbone.loc.3.bias', 'backbone.model.layer2.0.0.downsample.0.weight', 'backbone.model.layer2.0.0.bn2.running_mean', 'ConstantOfShape_682', 'backbone.model.layer2.0.2.bn2.bias', 'Constant_663', 'backbone.model.layer2.0.3.bn1.weight', 'backbone.model.layer2.0.0.bn1.bias', 'backbone.model.layer1.5.3.bn2.running_mean', 'backbone.model.layer1.5.3.conv2.weight', 'Constant_493', 'Constant_667', 'backbone.model.layer1.5.3.bn1.running_mean', 'Constant_673', 'backbone.loc.1.bias', 'backbone.model.layer1.5.3.bn1.bias', 'backbone.model.layer2.0.2.bn2.running_mean', 'Constant_665', 'backbone.model.layer1.5.3.bn1.weight', 'backbone.loc.2.bias', 'Constant_494', 'backbone.model.layer1.1.running_mean', 'Constant_541', 'backbone.model.layer2.0.2.bn1.running_mean', 'backbone.model.layer1.5.2.bn1.running_var', 'backbone.model.layer1.4.1.conv1.weight', 'backbone.additional_blocks.3.2.bias', 'backbone.model.layer2.0.0.downsample.1.running_mean', 'backbone.model.layer1.5.3.bn2.running_var', 'backbone.model.layer1.5.2.bn1.running_mean', 'Constant_523', 'backbone.model.layer1.5.1.bn1.bias', 'backbone.model.layer2.0.0.bn2.running_var', 'backbone.model.layer1.5.0.bn2.running_var', 'backbone.model.layer1.5.2.bn1.bias', 'backbone.additional_blocks.2.0.bias', 'backbone.model.layer2.0.2.bn1.bias', 'backbone.model.layer1.4.2.bn2.weight', 'Unsqueeze_445', 'backbone.model.layer1.4.1.bn1.running_var', 'backbone.model.layer2.0.0.bn1.running_mean', 'backbone.model.layer1.4.1.bn2.running_var', 'backbone.model.layer1.5.2.bn2.weight', 'backbone.conf.2.bias', 'Constant_620', 'backbone.model.layer1.5.0.downsample.1.running_var', 'backbone.model.layer1.4.2.bn1.weight', 'Constant_482', 'Constant_596', 'backbone.model.layer1.5.1.bn1.running_mean', 'Unsqueeze_444', 'Constant_670', 'backbone.model.layer1.5.3.bn2.weight', 'backbone.model.layer1.4.1.bn1.bias', 'Unsqueeze_467', 'backbone.model.layer2.0.0.bn2.weight', 'Constant_567', 'Constant_510', 'backbone.model.layer1.4.0.conv2.weight', 'backbone.model.layer1.4.0.bn2.running_mean', 'backbone.model.layer1.4.1.bn2.running_mean', 'backbone.model.layer1.5.2.bn2.bias', 'backbone.additional_blocks.2.0.weight', 'backbone.model.layer2.0.3.bn1.running_var', 'backbone.model.layer2.0.1.bn2.running_var', 'backbone.model.layer1.5.0.bn1.running_var', 'Constant_515', 'backbone.model.layer1.4.0.bn1.weight', 'backbone.model.layer1.4.0.bn2.running_var', 'backbone.model.layer1.5.3.bn2.bias', 'backbone.model.layer2.0.3.conv2.weight', 'backbone.model.layer1.5.2.conv2.weight', 'Constant_543', 'backbone.model.layer1.4.2.bn1.running_mean', 'backbone.model.layer2.0.1.bn1.running_mean', 'backbone.model.layer1.4.2.bn1.bias', 'backbone.model.layer1.5.1.bn1.running_var', 'backbone.model.layer1.1.running_var', 'backbone.model.layer2.0.5.bn1.running_var', 'Constant_480', 'Constant_607', 'backbone.model.layer1.4.1.conv2.weight', 'Constant_476', 'backbone.model.layer2.0.3.bn2.bias', 'backbone.model.layer2.0.0.conv2.weight', 'backbone.model.layer2.0.1.conv2.weight', 'backbone.model.layer2.0.1.bn1.bias', 'backbone.model.layer1.5.3.conv1.weight', 'backbone.model.layer1.4.1.bn2.bias', 'Unsqueeze_422', 'backbone.model.layer1.1.weight', 'backbone.model.layer2.0.2.bn2.weight', 'backbone.model.layer1.5.1.conv1.weight', 'backbone.model.layer1.4.0.conv1.weight', 'backbone.model.layer1.5.1.bn2.bias', 'Constant_625', 'backbone.model.layer1.1.bias', 'Constant_587', 'backbone.model.layer2.0.0.bn1.running_var', 'backbone.model.layer1.4.0.bn2.weight', 'backbone.model.layer2.0.2.bn2.running_var', 'backbone.model.layer1.4.0.bn2.bias', 'backbone.model.layer1.4.2.bn1.running_var', 'backbone.model.layer1.5.0.bn2.running_mean', 'backbone.model.layer1.5.3.bn1.running_var', 'backbone.model.layer1.5.0.downsample.1.running_mean', 'Constant_563', 'backbone.model.layer2.0.0.downsample.1.running_var', 'backbone.model.layer1.4.2.bn2.bias', 'Constant_487', 'Constant_533', 'backbone.model.layer2.0.4.bn1.weight', 'backbone.model.layer2.0.3.bn1.bias', 'backbone.model.layer1.4.1.bn1.running_mean', 'backbone.loc.0.weight', 'backbone.model.layer1.4.2.bn2.running_var', 'backbone.model.layer1.4.0.bn1.running_var', 'backbone.additional_blocks.0.0.weight', 'backbone.model.layer2.0.3.bn2.weight', 'backbone.model.layer2.0.0.bn2.bias', 'backbone.model.layer1.5.0.bn1.weight', 'backbone.model.layer1.5.2.bn1.weight', 'Constant_669', 'Constant_504', 'backbone.model.layer1.5.0.bn1.bias', 'backbone.model.layer1.5.0.bn1.running_mean', 'backbone.conf.3.bias', 'Constant_558', 'backbone.additional_blocks.1.0.weight', 'Constant_438', 'backbone.model.layer1.5.1.bn2.running_mean', 'backbone.model.layer1.5.0.conv2.weight', 'backbone.model.layer1.4.1.bn1.weight', 'backbone.model.layer1.4.1.bn2.weight', 'Unsqueeze_356', 'backbone.model.layer1.5.0.bn2.weight', 'Constant_474', 'backbone.model.layer1.5.1.conv2.weight', 'backbone.model.layer1.5.0.downsample.0.weight', 'backbone.model.layer2.0.5.conv2.weight', 'Constant_579', 'backbone.model.layer2.0.0.bn1.weight', 'backbone.model.layer1.5.0.downsample.1.bias', 'backbone.model.layer2.0.5.bn2.running_mean', 'backbone.model.layer1.5.0.conv1.weight', 'backbone.model.layer1.4.0.bn1.bias', 'Constant_526', 'backbone.model.layer2.0.5.bn1.weight', 'Constant_731', 'Unsqueeze_345', 'backbone.model.layer1.5.0.downsample.1.weight', 'backbone.model.layer1.5.1.bn2.weight', 'Constant_506', 'backbone.model.layer2.0.4.bn1.running_mean', 'Constant_632', 'Unsqueeze_346', 'backbone.model.layer2.0.4.bn1.running_var', 'backbone.model.layer2.0.0.downsample.1.weight', 'backbone.model.layer2.0.4.bn2.weight', 'backbone.model.layer2.0.4.bn2.bias', 'Constant_590', 'backbone.model.layer2.0.4.conv2.weight', 'backbone.model.layer2.0.4.bn2.running_mean', 'backbone.model.layer2.0.3.conv1.weight', 'backbone.model.layer1.5.1.bn2.running_var', 'backbone.model.layer1.5.2.conv1.weight', 'backbone.model.layer2.0.4.bn2.running_var', 'backbone.model.layer2.0.5.bn1.bias', 'Constant_552', 'backbone.model.layer2.0.5.bn1.running_mean', 'backbone.model.layer2.0.5.bn2.bias', 'backbone.model.layer2.0.5.bn2.running_var', 'backbone.additional_blocks.0.0.bias', 'backbone.additional_blocks.0.2.weight', 'Constant_532', 'backbone.additional_blocks.0.2.bias', 'backbone.loc.5.bias', 'backbone.additional_blocks.1.0.bias', 'Constant_668', 'Constant_503', 'Constant_555', 'backbone.additional_blocks.1.2.weight', 'Unsqueeze_411', 'backbone.additional_blocks.1.2.bias', 'Constant_664', 'backbone.additional_blocks.2.2.weight', 'Constant_491', 'Constant_633', 'Constant_405', 'backbone.additional_blocks.2.2.bias', 'Constant_611', 'Constant_460', 'backbone.additional_blocks.3.0.weight', 'Constant_637', 'backbone.additional_blocks.3.0.bias', 'backbone.additional_blocks.3.2.weight', 'backbone.model.layer2.0.1.bn2.weight', 'Unsqueeze_368', 'backbone.additional_blocks.4.0.weight', 'backbone.additional_blocks.4.0.bias', 'backbone.model.layer1.4.0.bn1.running_mean', 'Constant_485', 'backbone.additional_blocks.4.2.weight', 'Constant_687', 'Constant_572', 'backbone.additional_blocks.4.2.bias', 'backbone.conf.0.weight', 'backbone.loc.0.bias', 'Unsqueeze_466', 'Constant_652', 'Constant_339', 'Constant_674', 'backbone.conf.0.bias', 'Constant_516', 'Constant_350', 'Constant_521', 'backbone.conf.4.weight', 'backbone.model.layer1.5.2.bn2.running_mean', 'backbone.model.layer2.0.1.bn2.bias', 'Unsqueeze_357', 'Constant_557', 'backbone.loc.1.weight', 'Unsqueeze_367', 'backbone.conf.1.weight', 'Constant_586', 'backbone.loc.2.weight', 'backbone.model.layer1.4.2.conv1.weight', 'backbone.conf.1.bias', 'Unsqueeze_378', 'Constant_517', 'Constant_383', 'Constant_537', 'Unsqueeze_389', 'Constant_497', 'Unsqueeze_390', 'Constant_592', 'backbone.conf.2.weight', 'Constant_542', 'Constant_394', 'backbone.model.layer2.0.0.conv1.weight', 'Constant_502', 'Constant_628', 'Unsqueeze_400', 'backbone.model.layer2.0.0.downsample.1.bias', 'Unsqueeze_401', 'backbone.loc.3.weight', 'Constant_416', 'backbone.model.layer2.0.3.bn1.running_mean', 'Unsqueeze_423', 'backbone.loc.4.weight', 'Constant_513', 'Constant_372', 'backbone.loc.4.bias', 'Constant_626', 'backbone.model.layer1.5.1.bn1.weight', 'Constant_427', 'Constant_489', 'Unsqueeze_433', 'Constant_595', 'Constant_547', 'backbone.conf.3.weight', 'Unsqueeze_434', 'Constant_479', 'backbone.conf.4.bias', 'Constant_449', 'Constant_606', 'Unsqueeze_455', 'Unsqueeze_456', 'backbone.model.layer2.0.5.bn2.weight', 'Constant_496', 'backbone.conf.5.bias', 'Unsqueeze_379', 'Constant_475', 'backbone.model.layer2.0.2.bn1.running_var', 'backbone.model.layer2.0.5.conv1.weight', 'Constant_477', 'backbone.model.layer2.0.4.bn1.bias', 'Constant_481', 'Constant_361', 'Constant_484', 'Constant_486', 'Constant_646', 'Constant_492', 'Constant_550', 'Constant_498', 'Constant_499', 'Constant_508']) are missing from input feed (['image']).\")

YOLOv3 (ONNX Hub, opset=12)

For split_init pass (Checker output):
C.check_model_path(\\nonnx.onnx_cpp2py_export.checker.ValidationError: Field \\'type\\' of \\'value_info\\' is required but missing.\\n'

For split_predict pass (Checker output):
C.check_model_path(\\nonnx.onnx_cpp2py_export.checker.ValidationError: Field \\'elem_type\\' of \\'type\\' is required but missing.\\n'

YOLOv3 (ONNX Hub, opset=10)

For split_init pass (Checker output):
C.check_model_path(\\nonnx.onnx_cpp2py_export.checker.ValidationError: Field \\'type\\' of \\'value_info\\' is required but missing.\\n'

For split_predict pass (Checker output):
C.check_model_path(\\nonnx.onnx_cpp2py_export.checker.ValidationError: Nodes in a graph must be topologically sorted, however input \\'TFNodes/yolo_evaluation_layer_1/arange_5/delta:0\\' of node: \\nname: add__252 OpType: Add\\n is not output of any previous nodes.\\n\\n==> Context: Bad node spec for node. Name: TFNodes/yolo_evaluation_layer_1/arange_1__246_loop OpType: Loop\\n'

This behavior occurs on classification models as well (e.g., DenseNet-121-7 and Inception-2).

Start by reproducing the split_init and split_predict failures with the cited SSD and YOLOv3 ONNX Hub models at opsets 10 and 12. Inspect the generated model parts and run the ONNX checker; done means both passes produce valid nets whose required inputs are satisfied.