Difference output when running jetson nano and xavier AGX

[hoanganhtuanctes]

Hi all,
My environments:

• Python 3.7.
• Pytorch 1.9.0 (build from source)
• JetPack 4.5 (same images)

My case:

• Network: Object Detection using Pytorch: Backbone: resnet18: GitHub - MhLiao/DB: A PyTorch implementation of "Real-time Scene Text Detection with Differentiable Binarization".
• Same input, same source code (network)
• Run on Xavier AGX: correct output
• Run on Jetson Nano: wrong output (empty output)
  Tested on 2 devices (nano) with the same result.
• Checked:
  + Input to network: Correct.
  + Pytorch: run another network using resnet18 (classification): show correct output on jetson nano
  + Memory: enough, not overload

Who can support me in this case?
Thanks

[AastaLLL]

Hi,

Did you use the identical source to get the output?
More, may I know the JetPack version of Nano and Xavier?

Thanks.

[hoanganhtuanctes]

I have just updated. I found the reason but I could not explain why.
It come from DCN layer. I built it from same source code.

[AastaLLL]

I’m closing this topic due to there is no update from you for a period, assuming this issue was resolved.
If still need the support, please open a new topic. Thanks

Hi,

Could you share a simple reproducible sample so we can check this internally?
Thanks.