According to the docs Quickstart Guide — DeepStream 6.1.1 Release documentation the deepstream package 6.0.1 requires CUDA 11.4.
But TensorRT 8.0.1 requires CUDA 11.0 - 11.3, and there is no version of TensorRT for CUDA 11.4 (attached screenshot).
And right now in the docker image nvcr.io/nvidia/deepstream:6.0.1-devel there are two CUDA: 11.3 and 11.4, you can check it by the next command:
docker run -it nvcr.io/nvidia/deepstream:6.0.1-devel ls -l /usr/local/
drwxr-xr-x 1 root root 4096 Feb 18 23:42 bin
lrwxrwxrwx 1 root root 22 Sep 17 2021 cuda -> /etc/alternatives/cuda
lrwxrwxrwx 1 root root 25 Sep 17 2021 cuda-11 -> /etc/alternatives/cuda-11
drwxr-xr-x 6 root root 4096 Feb 18 22:42 cuda-11.3
drwxr-xr-x 1 root root 4096 Sep 21 2021 cuda-11.4
drwxr-xr-x 2 root root 4096 Aug 27 2021 etc
drwxr-xr-x 2 root root 4096 Aug 27 2021 games
drwxr-xr-x 1 root root 4096 Feb 18 23:51 include
drwxr-xr-x 1 root root 4096 Feb 18 23:51 lib
lrwxrwxrwx 1 root root 9 Aug 27 2021 man -> share/man
drwxr-xr-x 2 root root 4096 Aug 27 2021 sbin
drwxr-xr-x 1 root root 4096 Feb 18 23:42 share
drwxr-xr-x 2 root root 4096 Aug 27 2021 src
Please fix the next topics:
- fix base docker image
- fix doc . It said I must install CUDA 11.4 and TensorRT 8.0.1 but it is incompatible
- create a docker image with python3.8, because python3.6 is deprecated, has no support, and has many bags.
