Running on Jetson Xavier NX (aarch64)

Open shecker opened this issue 4 years ago • 1 comments

Hello

Thanks for this library. We have been trying for several days to get DCNv2_TRT running on an Nvidia Jetson Xavier NX, but unfortunately no luck.

We are getting a memory related error, but monitoring with jtop and/or tegra_stats reveals that GPU memory should not be a problem. Running on x86_64, we were able to verify that a maximum of 3.2GB are required and are able to run your code without issues, however on aarch64 we are seeing this issue and are a bit lost as to what to do next. The problem is not caused by inadequate workspace allocation. Have you been able to convert and run a model using DCNv2 on a Jetson family device and if so would you be able to share your approach?

Thanks a lot for your help!

Nov 29 '21 10:11 shecker

Have you solved this problem?

Dec 23 '24 08:12 1204Gin