cannnot install pytorch on tx2

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hello,I tried to install pytorch on tx2.First I followed the Install procedure for pyTorch on NVIDIA Jetson TX1/TX2 with JetPack <= 3.2.1. For JetPack 4.2 and Xavier/Nano/TX2, see https://devtalk.nvidia.com/default/topic/1049071/jetson-nano/pytorch-for-jetson-nano/ · GitHub However there are many erros,then i found
https://devtalk.nvidia.com/default/topic/1042821/jetson-tx2/pytorch-install-with-python3-broken/post/5290956/#5290956
and followed the commet6 of it,it errored as :
/home/nvidia/Tools/pytorch/torch/csrc/cuda/nccl.h:21:23: error: variable or field ‘throw_nccl_error’ declared void

I have set the use_nccl=false both in cmakelists and setup.py.Why does it stll make the nccl.h?

2

Hi,

Please clean build the package after updating.
Try this command first:

sudo python setup.py clean

Thanks.

3

Thanks for replying,but it doesn’t work

4

Hi, Im currently installing pytorch on Jetson TX2 myself. I had a lot of trouble on JetPack 3.1, so I opted to use JetPack 3.2. Im still finishing the first installation, so far no trouble.
If I succeed I will share my solution.
But if I may, did you $sudo nano CMakesList.txt and $sudo nano setup.py?
I`m having trouble setting NCCL off.

Thanks.

5

Hi,

We can reproduce this issue on our environment now.
Will update information with you once we find a solution.

Thanks.

6

I found a way around the problem and am currently working on a full tutorial -to be posted sometime this week-.

7

Finally, I was able to work my way around the problem.
The final tutorial is to be used after a clean JetPack flash.
I strongly suggest either JetPack3.2.1 or JetPack3.3

https://github.com/Bfzanchetta/DLARM/blob/master/jetsontx2/pytorch_install.sh

Have a good one!