Scott McKay
Scott McKay
Interestingly I can create a model programmatically that uses 'same_lower' for 'pad_type' and default 'pad' values and it will compile and run successfully. Naively that makes it seem like some...
The 'pad' certainly could be an artifact of Netron, however dumping the model spec by adding the below python to your script shows a 'pad' input exists. If I save...
Awesome. Thanks for the update. Is there a way to replicate things using the python builder code for the scenario where the coremltools python implicitly adds an input like the...
[This](https://github.com/apple/coremltools/issues/2127#issuecomment-1924906794) output was from adding the dumping logic to your builder based script. There's no 'pad' specified in the script, but it gets added to the model generated. I don't...
@chenfucn is this a known issue? Should we handle cpuinfo failing more gracefully? If it's not critical to have the cpu info maybe logging and ignoring the error is an...
It would be in the nightly package until the next official release. https://test.pypi.org/project/ort-nightly/
@jcampbell05 There has been no change that I can see to print a warning instead of failing with an exception. What error exactly are you seeing? https://github.com/microsoft/onnxruntime/blob/538d64891ac8e43c1faf7846635c3a1bf7b6b6c5/onnxruntime/core/common/cpuid_info.cc#L136
There's no exception thrown in the latest code so the failure is most likely coming from somewhere else. Problem is there's no default logger so the real error isn't clear....