tensorflow-onnx
tensorflow-onnx copied to clipboard
Remove the ml opset version that is saved by default without --extra_opset and --custom-ops
Remove ml version opset that is stored by default without using two flags(--extra_opset, --custom-ops)
This additional ml version opset can cause problems with converting for model serving for various edge devices
If additional ml version offset is required, add it via --extra_opset or --custom-ops
Directly removing it will cause other problem which has been covered by tests. Suggest to add a judgement around it so if extra_opset or custom-ops was specified, then skip this default value.