Questions about installing mxnet and gluoncv on nvidia jetson tx2

[Jessespace]

1. I tried the mxnet official website to use the source installation steps, failed after many attempts.
2. Use sudo pip install mxnet-jetson to install successfully, but the latest version is 1.2.1, and gluoncv requires mxnet version to be at least 1.3.0 (otherwise when importing gluoncv error)
   How do I upgrade the mxnet version on jetson to 1.3.0 or higher?

[mxnet-label-bot]

Hey, this is the MXNet Label Bot.
Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it.
Here are my recommended labels: Installation

[vdantu]

@mxnet-label-bot add [question, build]
@Jessespace : Could you please share the errors you faced when trying to compile from source?
@KellenSunderland : Would you be able to provide a context on mxnet-jetson package and its upgrade time frames?

[Jessespace]

@mxnet-label-bot add [question, build]
@Jessespace : Could you please share the errors you faced when trying to compile from source?
@KellenSunderland : Would you be able to provide a context on mxnet-jetson package and its upgrade time frames?

My device is nvidia jetson tx2

1. I followed the steps of the official website, failed. ( Sorry, I didn't record the errors. )
2. I referenced here to recompile and succeeded. This is the follow steps:

sudo apt-get update
sudo apt-get -y install git build-essential libatlas-base-dev libopencv-dev graphviz python-pip
sudo pip install pip --upgrade
sudo pip install setuptools numpy --upgrade
sudo pip install graphviz jupyter

git clone https://github.com/apache/incubator-mxnet.git --recursive
cd incubator-mxnet

cp make/config.mk .
sed -i 's/USE_CUDA = 0/USE_CUDA = 1/' config.mk
sed -i 's/USE_CUDA_PATH = NONE/USE_CUDA_PATH = \/usr\/local\/cuda/' config.mk
sed -i 's/USE_CUDNN = 0/USE_CUDNN = 1/' config.mk
sed -i '/USE_CUDNN/a CUDA_ARCH := -gencode arch=compute_62,code=sm_62' config.mk

Editing 3rdparty/mshadow/make/mshadow.mk and adding the following after line 122:
MSHADOW_CFLAGS += -DMSHADOW_USE_PASCAL=1

make -j $(nproc)

cd ..     # (incubator-mxnet dir)
export MXNET_HOME=$(pwd)
echo "export PYTHONPATH=$MXNET_HOME/python:$PYTHONPATH" >> ~/.bashrc
source ~/.bashrc

The main difference between the above steps and the official website tutorial is that:
cp make/config.mk cp make/crosscompile.jetson.mk config.mk ,What is the difference between the two?
In addition, can you provide compiled mxnet1.4.0 ? The compilation process is too long, Thanks.

[aaronmarkham]

@Jessespace Can you try the instructions here: https://github.com/apache/incubator-mxnet/blame/master/docs/install/install-jetson.md

[haojin2]

@Jessespace I've tried installation on Jetson TX2 before and I would definitely recommend using the cross-compilation method.