Created by Frank Schluenzen, last modified on Dec 06, 2018 10:41

mllab-amd01.desy.de is equipped with AMD EPYC 7351 (32-cores), 64GB ram and a Vega 10 Radeon Instinct MI25 GPU. This is a brief description of the setup for Ubuntu 18.04. A prior installation with Centos_7 was unsuccessful (GPU throws errors).

System installation

Following  instructions under https://rocm.github.io/ROCmInstall.html, https://github.com/ROCmSoftwarePlatform/tensorflow-upstream/blob/develop-upstream/rocm_docs/tensorflow-install-basic.md, 

wget -qO - http://repo.radeon.com/rocm/apt/debian/rocm.gpg.key | sudo apt-key add -
echo 'deb [arch=amd64] http://repo.radeon.com/rocm/apt/debian/ xenial main' | sudo tee /etc/apt/sources.list.d/rocm.list
apt update
apt install rocm-dkms
reboot

# check the GPU is properly recognized
dmesg | grep -i kfd
/opt/rocm/opencl/bin/x86_64/clinfo 
/opt/rocm/bin/rocminfo 

# add packages
apt install rocm-dkms rocm-dev rocm-libs rocm-device-libs hsa-ext-rocr-dev hsakmt-roct-dev hsa-rocr-dev rocm-opencl \
            rocm-opencl-dev rocm-utils rocm-profiler cxlactivitylogger miopen-hip miopengemm libnuma-dev
apt install build-essential clang clang-format clang-tidy cmake cmake-qt-gui g++-multilib libunwind-dev libfftw3-dev \
            libelf-dev libncurses5-dev libpthread-stubs0-dev gfortran libboost-program-options-dev libssl-dev libboost-dev \
            libboost-system-dev libboost-filesystem-dev rpm apt-utils pkg-config
apt install python-numpy  python-dev  python-wheel  python-mock  python-future  python-pip  python-yaml  python-setuptools
apt install python3-numpy python3-dev python3-wheel python3-mock python3-future python3-pip python3-yaml python3-setuptools

# users apparently need to be in group video to access the GPU
usermod -a -G video ...


Samples in user space

https://gpuopen.com/rocm-tensorflow-1-8-release/ has some basic tensorflow examples for using the AMD GPU. The system installation instructions on the page are outdated.

# check gpu access
/opt/rocm/bin/rocminfo                             

# check opencl
/opt/rocm/opencl/bin/x86_64/clinfo
wget https://raw.githubusercontent.com/bgaster/opencl-book-samples/master/src/Chapter_2/HelloWorld/HelloWorld.cpp
wget https://raw.githubusercontent.com/bgaster/opencl-book-samples/master/src/Chapter_2/HelloWorld/HelloWorld.cl
g++ -I /opt/rocm/opencl/include/ ./HelloWorld.cpp -o HelloWorld -L/opt/rocm/opencl/lib/x86_64 -lOpenCL
./HelloWorld

# tensorflow
pip3 install --user tensorflow-rocm
git clone https://github.com/tensorflow/models.git
cd ~/models/tutorials/image/imagenet
python3 classify_image.py

Links