TensorRT Python example code for reference | no directory for Python as per documentation

RatheeshR · May 10, 2019, 6:38pm

We are trying to check the TensorRT python example as per the documentation “https://docs.nvidia.com/deeplearning/sdk/tensorrt-sample-support-guide/index.html#uff_custom_plugin”.

As per the documentation, the sample will be installed in the “/usr/src/tensorrt/samples/python/uff_custom_plugin” directory.

But We do not have the ‘python’ directory in the ‘/usr/src/tensorrt/samples/’, for Drive PX2, with Driveworks_1_2 (Tegra B re-flashed Today) .

Can you share some examples of the TensorRT with Python and tensorflow object detection, for reference.

AastaLLL · May 13, 2019, 3:56am

Hi,

The document you checked is for TensorRT 5.1(latest) but PX2 is still using TensorRT4.1.
Python sample is not available on the PX2 ~~yet~~.
~~The Python support for Tegra starts from version 5.0.~~

Thanks.

RatheeshR · May 13, 2019, 4:01am

can you share an example for python +tensorRT
When will PX2 get TensorRT 5.x version?

SivaRamaKrishnaNV · May 13, 2019, 5:27am

Dear RatheeshR,
There is no more releases targeted on Drive PX2. Just to clarify, python support from TRT 5.x available on only Jetson platform.So The workflow in your case would be like below.

convert your tensorflow model to UFF on Host(Refer https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt_401/tensorrt-developer-guide/index.html#working_tf)
copy UFF model onto Drive PX2 and get TensorRT model from Drive PX2 using C++ API(Refer Sample https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt_401/tensorrt-developer-guide/index.html#mnist_uff_sample)
Perform inference using TensorRT model

As there are no releases planned on Drive PX2, please consider upgrade to Drive AGX platform to get new features on TRT and DW.