Hello everyone.
For some reason, I use jetson nano to run our computer vision product (using deepstream).
However, after installing two components (host components, Jetson SDK components). My storage is full.
Hi @DaneLLL . Thank you for your help.
I followed and installed all packages in the link that you mentioned.
However, after pulling nvcr.io/nvidia/deepstream-l4t:5.1-21.02-samples from nvidia-ngc. I cant run my code - (runs normally on jetson nano sd card)
The problem is I can’t create pgie or any deepstream-plugin.
Here ís the log:
Creating Pipeline
Creating streammux
Creating uridecodebin for [rtsp://admin:abcd1234@192.168.1.104:554/Streaming/Channels/101?transportmode=mcast&profile=Profile_1]
source-bin-00
Creating Pgie
Unable to create pgie
Exception in thread Inference-Thread:
Traceback (most recent call last):
File "/usr/lib/python3.6/threading.py", line 916, in _bootstrap_inner
self.run()
File "/usr/lib/python3.6/threading.py", line 864, in run
self._target(*self._args, **self._kwargs)
File "/ws/intrusion_detections/streaming/app/libs/detector_stream.py", line 613, in run_first_stream
create_pipeline(uri_name)
File "/ws/intrusion_detections/streaming/app/libs/detector_stream.py", line 360, in create_pipeline
pgie_batch_size = pgie.get_property("batch-size")
AttributeError: 'NoneType' object has no attribute 'get_property'
Here is the log when I run default deepstream in docker.
deepstream-app: error while loading shared libraries: /usr/lib/aarch64-linux-gnu/libnvinfer.so.7: file too short
Hi @foreverneilyoung , I mention you because your topic is closed.
I also had the same problem as you. @DaneLLL helped me to run deepstream successfully on jetson nano os (ubuntu). However, we are not able to run with docker (5.1-l4t).
Have you ever encountered this problem? can you help me?
The problem seems to be when you move /usr/local to somewhere else, e.g. the sd-card:
I’ve had to move some nvidia stuff to from /opt/nvidia to /media/sd-card/opt-nvidia/ due to emmc space problems. Same for /usr/local to /media/sd-card/usr-local/
The NVIDIA docker runtime seems to do something weird, where it actually mounts some library directories into the docker. So actually some dependencies are outside (the opposite of what docker is supposed to be…)
It would seem that /usr/local would normally mapped to be inside the container, but the container now maps /media/sd-card/usr-local/ instead.
This means that /usr/local isn’t in the container and builds fail. You need to therefore add e.g. “-L/media/sd-card/usr-local/cuda-10.2/targets/aarch64-linux/lib/” in the makefiles etc for it to find libraries, such as libcublas which is now accessable at /media/sd-card/usr-local/cuda-10.2/targets/aarch64-linux/lib/libcublas.so
It also means you need to do e.g. " export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/media/sd-card/usr-local/cuda-10.2/targets/aarch64-linux/lib/ " to allow libraries to be found at runtime