TRT Dynamic shape for UFF format

I cann’t apply dynamic shape to UFF models.

I should call “registerInput(inputName, inputDims, inputOrder)” to parse UFF file by developer guide.
If so, I can not use dynamic shape because the prediction engine has the fixed input dimension.

So I wonder that I MUST use ONNX to get dynamic shape or not.

Dynamic shape is currently not supported by UFF parser.

You can try tf2onnx + ONNX parser as an alternative.


Dear SunilJB,
Thanks for your answer.

I am trying to tf2onnx, but I have a few issues to convert and parse.
Anyway, I will solve them!

Thanks again.