Jetson-inference compiler fails at [57%]

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

Hi there,

I am using a 4 Gb nano with B01 carrier board. 64 Gb SD card. Headed and following the S3E2 tutorial and working in the docker container.

Under ‘Compiling the project’ I ran ldconfig (omitting $ sudo)
Compiling proceeds until
‘[57%] Linking CXX shared library aarcg64/lib/libjetson-inference.so’

next line:

‘/usr/bin/ld: cannot find - lnvcaffe_parser’

and from there follows a series of error and failure messages.

Can you offer any advice? Up to this point everything went 100% as per the AI tutorial.

Thank you

2

Hi @pddarrell, hmm interesting. Which version of JetPack are you using? Can you try removing -lnvcaffe_parser from this line of the CMakeLists.txt?

3

Hi Dusty.

Great tutorials, by the way.

JetPack version r32.5.0

In the CMakeLists.txt, do I remove _plugin nvcaffe_parser from line 169?
The line currently reads 'target_link_librariies(jetson-inference jetson-utils nvinfer nvinfer_plugin nvcaffe_parser)

Do I then run the whole compiler section again?

Thanks for your help.

4

Yep, try removing nvcaffe_parser from that line. Then re-run cmake ../ and make

5

I removed nvcaffe_parser from line 169, saved the .txt, and re-ran make.

at [ 49%] it returns:
/usr/bin/ld: cannot find -lnvcaffe_parser
collect2: error: ld returned 1 exit status

(etc.)

please advise

6

I have currently replaced it, saved, reran cmake …/ and make

same error

7

Are you sure that your jetson-inference/CMakeLists.txt no longer has -lnvcaffe_parser?

If you grep -r -i nvcaffe_parser /path/to/your/jetson-inference does it show nvcaffe_parser in the CMakeLists.txt?

8

Also, if you continue to have issues building from source, you can try using the pre-compiled jetson-inference docker container instead:

9

In reverse order:
“Also, if you continue to have issues building from source, you can try using the pre-compiled jetson-inference docker container instead:”

As per first post, I am working in the docker container.

“Are you sure that your jetson-inference/CMakeLists.txt no longer has -lnvcaffe_parser ?”

grep shows nvcaffe_parser in the CMakeLists.txt however I have rebooted and the line does not appear in the .txt in the gedit text editor. I have no way of removing it. (do I?)

please advise

10

Sorry, I missed the part where you are working in the docker container. If you are already in the container, the project is already built and you don’t need to build it again.

If you want to re-build the docker container, make you changes to your source try outside of the container, and then see here:

11

I think the problem may be down to the fact that I had been checking the earlier classification course (S1E3) and followed the video link to S3E3. I have started from the beginning again (using the pre-compiled docker) and have got to ‘using the ImageNet program on Jetson’.

I am in nvdli-nano/jetson-inference/build/aarch64/bin

When I run:
./imagenet.py images/orange_0.jpg images/test/output_0.jpg

I get a Traceback:
File “./imagenet.py”, line 24, in
import jetson.inference
ModuleNotFoundError: No module named ‘jetson’

Am I missing a file?
Should I open this as a new case?

Thank you

12

Case is solved.

I hadn’t used 'docker/run.sh

Thank you for your assistance Rusty.

13

OK gotcha, great - glad you got it working. Yes, the jetson-inference projects runs in a different container than the nvdli container.

14

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.