Issues with libnvfatbinaryloader lib

[guy]

Hello to everyone,

Working on a Jetson NX/Nano/DGPU
Deepstream Version is 5.1
Jetpack on jetsons is 4.5
TensorRT on DGPU:7.0.0 and on the Jetsons:7.1.0
Nvidia GPU driver on the DGPU is:
465 and cuda version is:11.3
Using the following docker images:
nvcr.io/nvidia/deepstream-l4t:5.1-21.02-samples
nvcr.io/nvidia/deepstream-l4t:5.1-21.02-iot
nvcr.io/nvidia/deepstream-l4t:5.1-21.02-base
for jetsons.
and:
nvcr.io/nvidia/deepstream:5.1-21.02-base
nvcr.io/nvidia/deepstream:5.1-21.02-devel
nvcr.io/nvidia/deepstream:5.1-21.02-samples
nvcr.io/nvidia/deepstream:5.1-21.02-iot
for the DGPU.
When Trying to load a pipeline with libnvidia-fatbinaryloader.so.440.18 for multi batch related jobs i got the following libnvidia-fatbinaryloader.so.440.18: file too short

Went over to /usr/lib/aarch64-linux-gnu in the container and saw that the filesize for
libnvidia-fatbinaryloader.so.440.18 is 0.
Saw another file (probably the same one just different name):
libnvidia-fatbinaryloader.so.32.4.4
so i’ve symlinked it to libnvidia-fatbinaryloader.so.440.18, but now deepstream refuse to load.

Any help would be greatly appreciated.

Thank you.

[bcao]

Can you refer Quickstart Guide — DeepStream 6.1.1 Release documentation

[guy]

I did and i’ve installed the jetson accordingly,but as a quick side-note, i’m running this from the docker image and not mounting anything from the os.

[bcao]

Could you reflash the latest Jetpack and see if the issue persist, I think the issue is caused by the setup.