NVIDIA Drive PX2 : Run pytorch model directly on board without tensorrt

Dear Team,

Is it possible to run any model(Pytorch,onnx,TF,caffe) directly on Drive PX2 target device without optimizing through tensorRT?

Is it possible to setup Anaconda environment on Tegra Linux OS to run the model directly on target GPUs?

Is it possible to upgrade the TensorRT version on Drive PX2?. currently SDK manager has installed the TensorRT 4.0.1.

Kindly let me know the process and do the needful.

Thanks and Regards,
Vyom Mishra

Dear @vyom.mishra,
There are no releases targetted for DRIVE PX2 now and last release has TRT 4.0.1.

We have not tested setting up anaconda. Do you see any issues in setting up?

You can install TF/pytorch natively and run models witbout TensorRT. However, we recommend using TensorRT for better performance on DRIVE PX2.

Dear Sir,

I tried settting up the anaconda and pip virtual environment on the Drive PX 2 OS linux.

When I tried to install the modules(through requirements.txt) , I face below error always:

ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output
I have attached virtualenv setup issues in the txt file for your reference:
pip_virtualenv_drivepx2.txt (39.4 KB)

Do let me know if we can resolve the same.

Thanks and Regards,
Vyom Mishra

Hi @vyom.mishra, here is the compilation/installation guide for PyTorch (1.7.1) on Drive PX2. Hope it helps.

Vishal Balaji

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.