Deepstream with python bindings

So the Warning is gone but the cuda issue is back.

Immediately after installing the system with DS7 via SDK Manager it looks like it has not completed the installation. I have to install firefox to get browser. Strange behavior compared to previous releases.

  1. SDK Manager install completes
  2. Installing dependencies, but change Python3.8-dev to 3.10-dev
  3. Install PYDS for 1.1.11 version with the whl.
  4. sudo git clone GitHub - NVIDIA-AI-IOT/deepstream_python_apps: DeepStream SDK Python bindings and sample applications -b v1.1.11
    to: /opt/nvidia/deepstream/deepstream-7.0/sources

Then ran the …/apps/deepstream-test1/deepstream_test1.py.
This produces the same error.

I then installed cuda-python on the system.

pip3 install cuda-python

This did not do the trick. Uninstalled it with

pip3 uninstall cuda-python

Then removed deepstream_python_apps from /opt…/sources
Git cloned the latest to /opt/…

sudo git clone https://github.com/NVIDIA-AI-IOT/deepstream_python_apps.git

Started the application and it produced a lot of Warnings and had to convert to int8 etc but after a while it did start.

I don’t know why this works or if something is not ok. But I will have to install torch and torchvision too.
I will reinstall now again from scratch and go with the latest deepstream-python-apps from the start and not
specify the version with -b x.x.x. Lets see. Keep you posted.

For torch I will use 2.1.0
https://nvidia.box.com/shared/static/0h6tk4msrl9xz3evft9t0mpwwwkw7a32.whl
For torchvision v0.16.1

git clone --branch <version> https://github.com/pytorch/vision torchvision   # see below for version of torchvision to download
$ cd torchvision
$ export BUILD_VERSION=0.x.0  # where 0.x.0 is the torchvision version  
$ python3 setup.py install --user