Issue with onnx opset converters

[pddarrell]

Please provide the following information when requesting support.

• Hardware (RGX 3080)
• Network Type (Classification)
• TLT Version format_version: 2.0
toolkit_version: 3.22.05
published_date: 05/25/2022()
• Training spec file
• How to reproduce the issue ?

Hi

I intend to retrain this ResNet50 model using BYOM Converter:

It is opset 7 and the documentation offers the following on opset versions that BYOM Converter has been tested on:

" TAO BYOM Converter has been tested on opset version from 9 to 13. If your opset version is either lower than 9 or higher than 13, run this version conversion script from ONNX to convert the ONNX model to an appropriate version."
Following the link shown in the documentation leads to the following list of opset converters:

none of them convert from outside the range of opset 9 to 13 to within that range.
Do you have answers to these questions please:

1. Why was BYOM converter only tested on opsets that have no opset converters?
2. Why does the documentation direct to a site where no solution to the opset issue is actually offered?
3. Does this mean that I cannot use the ResNet50 .onnx model?

[Morganh]

There is no update from you for a period, assuming this is not an issue anymore.
Hence we are closing this topic. If need further support, please open a new one.
Thanks

As mentioned in the user guide,
TAO BYOM Converter has been tested on opset version from 9 to 13. If your opset version is either lower than 9 or higher than 13, run this version conversion script from ONNX to convert the ONNX model to an appropriate version.

Then, the Python API is described, with example, here.
You can convert the ONNX model to an appropriate version.

BTW, some examples can be found in Problem with version_converter · Issue #2072 · onnx/onnx · GitHub or Version converter 9-->10 fails with 'Input conv_first.weight is undefined!' · Issue #2995 · onnx/onnx · GitHub .

@pddarrell

Does this mean that I cannot use the ResNet50 .onnx model?

You can use the ResNet50 .onnx model. Need to convert Opset version to 9~13.