Dmitri Smirnov
Dmitri Smirnov
@hanzhaogang Please, try following the most recent updated instructions for building on Windows.
Which backend are you using to run this model? perhaps you can file against ONNX Runtime or Pytorch?
You might get a better response at github.com/pytorch/pytorch
FYI: https://github.com/microsoft/onnxruntime/pull/13215 perhaps, you could hold off on this?
Closing this due to clang-tidy pollution
I do not think we should expose them as a separate entity. This is not a supportable pubic interface. We use it for optimization purposes and some other internal needs....
What API do you have in mind? You can use BFloat16 with C/C++/C# API.
One question. Does this code still copy memory for input/output? We should not copy at least for input and if so I believe there is a way to prevent that...
Well, we are doing just find in C#, holding to the native pointers received from ORT and destroy then when IDisposable is invoked, similar to AutoClosable. If that does not...
/azp run Linux CPU CI Pipeline, Linux CPU Minimal Build E2E CI Pipeline, Linux CPU x64 NoContribops CI Pipeline, Linux GPU CI Pipeline, Linux GPU TensorRT CI Pipeline, Linux Nuphar...