This wasn't clear from the docs or the API types. In Session::runInputTensor and OutputTensor are used for inference but these only contain 1D values and no way to provide a shape for the tensor. If I have dynamic axes (batch size, sequence length etc). How can I specify the shape and also for the output get the output shape (which is also dynamic)?
This wasn't clear from the docs or the API types. In
Session::run
InputTensor
andOutputTensor
are used for inference but these only contain 1D values and no way to provide a shape for the tensor. If I have dynamic axes (batch size, sequence length etc). How can I specify the shape and also for the output get the output shape (which is also dynamic)?