Framework is not detected correctly from model format

[Charles-258]

The imput model format is onnx, and the framework onnxruntime can not detect the model format.

[mengniwang95]

Hi, could you provide your launcher codes and installed pkg to help locate problems?

[Charles-258]

Hi, could you provide your launcher codes and installed pkg to help locate problems?

Here is the link of code and data: https://drive.google.com/file/d/1ep0jKgnodvN8mL1E6Fsr08B2T5qCkX46/view?usp=sharing and thanks for your help.

[mengniwang95]

Do you install onnxruntime-extensions?

[Charles-258]

Do you install onnxruntime-extensions?

I have just install this package, but it still cant get the proper quantitized model.

[mengniwang95]

Sorry for my late reply, could you paste your error info? I try your script and it shows INC can detect model correctly but the input format is incorrect.

[Charles-258]

Here is my error info. Thanks for your help.

[mengniwang95]

This bug is caused by inference time collection. Since you don't provide eval_func or eval_dataloader+metric, it generates a fake eval_func automatically and it seems to jump in and jump out this eval_func at the same time... you can change line137 of objective.py to self.duration >= 0 locally to fix this problem and we will fix it in next release.