Inconsistent and incorrect use of operator descriptor in onnx parsing

[nives-vukovic]

There are onnx parse() functions which use opd.op_name inside MIGRAPHX_THROW macro, although this operator descriptor field is not initialized.

List of such files:

• src/onnx/parse_instancenorm.cpp
• src/onnx/parse_randomnormal_ops.cpp
• src/onnx/parse_randomuniform_ops.cpp
• src/onnx/parse_resize.cpp

In src/onnx/parse_qlinearpooling.cpp parse() function opd.onnx_name is used inside MIGRAPHX_THROW macros, which is not consistent with other prints (However, could be a better solution as onnx_name is more precise and op_name sometimes does not point out precisely which onnx operator causes an error).

There are onnx parse() functions which use opd.op_name inside tune_axis function, although this operator descriptor field is not initialized.

List of such files:

• src/onnx/parse_dequantizelinear.cpp
• src/onnx/parse_gather_elements.cpp
• src/onnx/parse_onehot.cpp
• src/onnx/parse_quantizelinear.cpp
• src/onnx/parse_split.cpp
• src/onnx/parse_unique.cpp

[sohaibnd]

Hi @nives-vukovic, thanks for pointing this out. opd.op_name has been replaced with opd.onnx_name inside MIGRAPHX_THROW macros (and tune_axis) since onnx_name is more descriptive as you've mentioned already. See the following PR: #3532

I am going to close this issue as it is resolved. If you come across any other problems, feel free to create another issue.