question-mark
Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Build fails due CUDNN (Xavier NX, JP 4.4.1) - ODC Darknet fork getting old?

See original GitHub issue

Greetings,

I have made a fresh install of Jetpack, and I notice the Darknet fork provided by Opendatacam fails to build in this configuration. Make file settings:

GPU=1
CUDNN=1
CUDNN_HALF=0
OPENCV=1
AVX=0
OPENMP=0
LIBSO=0
ZED_CAMERA=1 # ZED SDK 3.0 and above
ZED_CAMERA_v2_8=0 # ZED SDK 2.X

This is the error message:

./src/convolutional_layer.c: In function ‘cudnn_convolutional_setup’:
./src/convolutional_layer.c:286:24: error: ‘CUDNN_CONVOLUTION_FWD_PREFER_FASTEST’ undeclared (first use in this function); did you mean ‘CUDNN_CONVOLUTION_BWD_FILTER_ALGO_3’?
     int forward_algo = CUDNN_CONVOLUTION_FWD_PREFER_FASTEST;
                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                        CUDNN_CONVOLUTION_BWD_FILTER_ALGO_3
compilation terminated due to -Wfatal-errors.
Makefile:162: recipe for target 'obj/convolutional_layer.o' failed
make: *** [obj/convolutional_layer.o] Error 1

I have also tried with CUDNN_HALF on 0, and I get the exact same build error.

When I build darknet directly from the git, it builds just fine with the exact same configuration. It seems the ODC fork is getting a bit older now, it might be something in the latest jetpack which does not work well with that. Would it be possible to update the ODC darknet or will that break things?

Issue Analytics

  • State:closed
  • Created 3 years ago
  • Comments:13 (6 by maintainers)

github_iconTop GitHub Comments

3reactions
tdurandcommented, Jan 7, 2021

Hi @Denizzje

Yes, sorry, the latest jetpack doesn’t play well with our darknet fork version, but good news is that the “latest” darknet should work out of the box: https://github.com/alexeyab/darknet

The diff of our fork is just a few comments to clean up the logs: https://github.com/opendatacam/darknet/pull/7/files

I’m aiming to spend time later this months to update OpenDataCam docker images + the fork in order to be compatible with latest jetpack… In the meantime if you try it would be nice if you can let me know if it works

Thanks

1reaction
vsawcommented, Jan 9, 2021

Yes. Building Darknet takes 5-7 minutes which we could shave off the docker build. So this repo becomes more a versioned package builder then a “source store”.

Read more comments on GitHub >

github_iconTop Results From Across the Web

Darknet compile error with cudnn 8 - Jetson Xavier NX
Hi all, Do you know how to compile YOLOv4 on NX? i got this error with cudnn 8 “CUDNN_CONVOLUTION_FWD_PREFER_FASTEST” was not declared.
Read more >
A brand new website interface for an even better experience!
Build fails due CUDNN (Xavier NX, JP 4.4.1) - ODC Darknet fork getting old?
Read more >
Error build darknet windows 10 cuda and cudnn 11.2
I want to build darknet for YOLOv4. my configuration is: win10, microsoft visual studio code 2019, openCV 4.5.1, cuda and cudnn 11.2.
Read more >
YOLOv4 on Jetson Nano - JK Jung's blog
Basically I just cloned the darknet code from GitHub and followed the instructions in the 2. How to compile on Linux -> Using...
Read more >
How to install Yolo v4 on Jetson Xavier NX - Pysource
In this tutorial, we will see how to use Jetson Xavier NX with YOLO v4 and darknet. In the past, we have already...
Read more >

github_iconTop Related Medium Post

No results found

github_iconTop Related StackOverflow Question

No results found

github_iconTroubleshoot Live Code

Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start Free

github_iconTop Related Reddit Thread

No results found

github_iconTop Related Hackernoon Post

No results found

github_iconTop Related Tweet

No results found

github_iconTop Related Dev.to Post

No results found

github_iconTop Related Hashnode Post

No results found