The objective is to give you clear instruction on how to install packages in ARM platform, especially in Jetson family. This instruction was done under Python 3. Tests have been made on Jetson TX2 and Jetson Xavier. You may change sudo python3 and sudo pip3 into sudo python2 and sudo pip, respectively, to make it work under Python 2.
Before performing any installations, you may need to install the basic dependencies first.
$ sudo apt-get install cmake
$ sudo apt-get install python3-pip
$ sudo pip3 install wget
$ sudo pip3 install Cython
$ sudo apt-get install libboost-all-dev
$ sudo apt-get install python-numpy
$ sudo apt-get install build-essential python-dev python-setuptools libboost-python-dev libboost-thread-dev
You need to download PyCUDA from https://pypi.org/project/pycuda/#files. In the same directory of your PyCUDA download, run this terminal
$ tar xzvf pycuda-VERSION.tar.gz
$ cd pycuda-VERSION
Open configure.py and change the /usr/bin/env python into /usr/bin/env python3
$ ./configure.py
$ make -j4
$ sudo python3 setup.py install
$ sudo pip3 install .
In this tutorial, I used LLVM 7.0.1.
$ wget http://releases.llvm.org/7.0.1/llvm-7.0.1.src.tar.xz
$ tar -xvf llvm-7.0.1.src.tar.xz
$ cd llvm-7.0.1.src
$ mkdir llvm_build_dir
$ cd llvm_build_dir/
$ cmake ../ -DCMAKE_BUILD_TYPE=Release -DLLVM_TARGETS_TO_BUILD="ARM;X86;AArch64"
$ make -j4
$ sudo make install
$ cd bin/
$ echo "export LLVM_CONFIG=\""`pwd`"/llvm-config\"" >> ~/.bashrc
$ echo "alias llvm='"`pwd`"/llvm-lit'" >> ~/.bashrc
$ source ~/.bashrc
$ sudo pip3 install llvmlite==0.30.0
$ wget https://github.com/llvm/llvm-project/releases/download/llvmorg-10.0.1/llvm-10.0.1.src.tar.xz
$ tar -xvf llvm-10.0.1.src.tar.xz
$ cd llvm-10.0.1.src.tar.xz
$ mkdir llvm_build_dir
$ cd llvm_build_dir/
$ cmake ../ -DCMAKE_BUILD_TYPE=Release -DLLVM_TARGETS_TO_BUILD="ARM;X86;AArch64"
$ make -j4
$ sudo make install
$ cd bin/
$ echo "export LLVM_CONFIG=\""`pwd`"/llvm-config\"" >> ~/.bashrc
$ echo "alias llvm='"`pwd`"/llvm-lit'" >> ~/.bashrc
$ source ~/.bashrc
$ cd
$ git clone https://github.com/wjakob/tbb.git
$ cd tbb/build
$ cmake ..
$ make -j
$ sudo make install
$ pip3 install llvmlite
$ pip3 install numba
You may install Protobuf directly from Python using pip:
$ sudo pip3 install protobuf
If that doesn't work, you may build from source:
$ git clone https://github.com/protocolbuffers/protobuf.git
$ cd protobuf
$ git submodule update --init --recursive
$ ./autogen.sh
$ ./configure
$ make -j4
$ make check
$ sudo make install
$ sudo ldconfig
Before proceeding to the ONNX installation, you need to perform Protobuf installation above first since ONNX heavily relies on Protobuf installation.
$ sudo pip3 install onnx
$ sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev
$ sudo apt-get install libblas-dev liblapack-dev libatlas-base-dev gfortran
$ sudo pip3 install scipy
$ sudo pip3 install keras
$ sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev
$ sudo pip3 install -U numpy==1.16.1 future==0.17.1 mock==3.0.5 h5py==2.9.0 keras_preprocessing==1.0.5 keras_applications==1.0.6 enum34 futures testresources setuptools protobuf
If you have trouble installing Protobuf, you may take a look at the Protobuf installation part above. If everything's fine, you may proceed. Now, take a look at this https://developer.download.nvidia.com/compute/redist/jp/. You may find a lot of JetPack versions and you need to choose one based on your preference. In this tutorial, I used v411 since it has compatibility with Python 2.7 and Python 3.6 installation.
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v411 tensorflow-gpu
If you run into a problem when execute the code above like this:
Exception:
Traceback (most recent call last):
...
File "/usr/share/python-wheels/requests-2.18.4-py2.py3-none-any.whl/requests/models.py", line 935, in raise_for_status
raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 404 Client Error: Not Found for url: https://developer.download.nvidia.com/compute/redist/jp/v411/grpcio/
then you should download the Tensorflow installation manually. Previously, when I executed this sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v411 tensorflow-gpu, it already collected tensorflow-gpu and showed an output like this:
Collecting tensorflow-gpu
Downloading https://developer.download.nvidia.com/compute/redist/jp/v411/tensorflow-gpu/tensorflow_gpu-1.13.0rc0+nv19.2-cp36-cp36m-linux_aarch64.whl (204.6MB)
100% |████████████████████████████████| 204.6MB 4.6kB/s
So, I downloaded tensorflow_gpu-1.13.0rc0+nv19.2-cp36-cp36m-linux_aarch64.whl. From the download directory, I ran this inside the terminal:
$ sudo pip3 install tensorflow_gpu-1.13.0rc0+nv19.2-cp36-cp36m-linux_aarch64.whl
#2 Thank you for surefyyq for the solution regarding the TBB issue.
$ git clone https://github.com/wjakob/tbb.git
$ cd tbb/build
$ cmake ..
$ make -j
$ sudo make install