Ask a Question
Question
I used tf2onnx to convert a tensorflow to onnx, it converted and used normally.
Then I updated some packages like tensorflow, protobuf, after that I use tf2onnx to convert the same original tensorflow again, everything looks well.
Then I use polygraphy run to compare the result of the two onnx models, it outputs many differences.
My question is: is the differences a bug, how the differences caused in details.
Thanks!
Further information
- Is this issue related to a specific model?
Model name:
Model opset:
Notes