Install Tensorflow 2 for AMD GPUs by Eric Ngo Analytics Vidhya Medium PDF

Preview

Summary

instalando tensorflow...

Description

6/9/2021

Install Tensorflow 2 for AMD GPUs | by Eric Ngo | Analytics Vidhya | Medium

This is your last free member-only story this month. Sign up for Medium and get an extra one

Install Tensorflow 2 for AMD GPUs Eric Ngo

Follow

Dec 6, 2020 · 3 min read

AMD has released ROCm, a Deep Learning driver to run Tensorflow-written scripts on AMD GPUs. However, many owners and I have encountered many challenges in installing Tensorflow on AMD GPUs. Hence, I provided the installation instructions of Tensorflow for AMD GPUs below.

In this post, I guided you to install ROCm (AMD GPU driver) and Tensorflow ROCm (version compatible with AMD hardware) from Debian repository on Ubuntu. At the https://medium.com/analytics-vidhya/install-tensorflow-2-for-amd-gpus-87e8d7aeb812

1/6

6/9/2021

Install Tensorflow 2 for AMD GPUs | by Eric Ngo | Analytics Vidhya | Medium

moment this post published, ROCm runs only on Linux-like platforms (e.g. Ubuntu, Centos). To run Deep Learning with AMD GPUs on MacOS, you can use PlaidML owned and maintained by PlaidML. So far, I have not seen packages to run AMD-based Deep Learning on Windows. Updated on March. 1, 2021

Installation steps: Install GPU driver, ROCm Install AMD-compatible Tensorflow version, Tensorflow ROCm Notice: there is often a version mismatch between ROCm and TF-ROCm, and therefore, please check this link to find the compatible ROCm and TF-ROCm versions.

Prerequisites Operating Systems: Ubuntu 18.04.5 (Kernel 5.4) and other supported OSs. Hardware: Radeon VII (I owned it) and other supported hardware (e.g. Vega 64, RX580).

Install ROCm Notes: to install ROCm successfully, ensure the fresh installation (aka OS reboot) or clearly remove ROCm-related packages. Update the OS

sudo apt update sudo apt dist-upgrade sudo apt install libnuma-dev sudo reboot

https://medium.com/analytics-vidhya/install-tensorflow-2-for-amd-gpus-87e8d7aeb812

2/6

6/9/2021

Install Tensorflow 2 for AMD GPUs | by Eric Ngo | Analytics Vidhya | Medium

Add the ROCm apt repository The next code block is to install the latest ROCm version.

wget -q -O - https://repo.radeon.com/rocm/rocm.gpg.key | sudo apt-key add echo 'deb [arch=amd64] https://repo.radeon.com/rocm/apt/debian/ xenial main' | sudo tee /etc/apt/sources.list.d/rocm.list

ROCm is often updated before the Tensorflow ROCm. Hence, there is usually the version mismatch between ROCm and Tensorflow ROCm. The version-mismatch error code is “stack smashing detected”. To solve the problem, you can install older ROCm versions by slightly modifying the above code block. Also, check this link to find the ROCm version matched for the Tensorflow 2.x.x. version you want to install.

wget -q -O - https://repo.radeon.com/rocm/apt/x.x.x/rocm.gpg.key | sudo apt-key add echo 'deb [arch=amd64] https://repo.radeon.com/rocm/apt/x.x.x/ xenial main' | sudo tee /etc/apt/sources.list.d/rocm.list

that x.x.x is replaced by the ROCm version number. For example, x.x.x is replaced by 3.9.1 or x.x.x is replaced by 3.9 (0 is degraded from 3.9.0) Install ROCm meta-packages

sudo apt update sudo apt install rocm-dkms && sudo reboot

Add users to video & render groups

sudo usermod -a -G video $LOGNAME sudo usermod -a -G render $LOGNAME https://medium.com/analytics-vidhya/install-tensorflow-2-for-amd-gpus-87e8d7aeb812

3/6

6/9/2021

Install Tensorflow 2 for AMD GPUs | by Eric Ngo | Analytics Vidhya | Medium

echo 'ADD_EXTRA_GROUPS=1' | sudo tee -a /etc/adduser.conf echo 'EXTRA_GROUPS=video' | sudo tee -a /etc/adduser.conf echo 'EXTRA_GROUPS=render' | sudo tee -a /etc/adduser.conf sudo reboot

Notes: ‘render’ is only supported on Ubuntu 20.x.x and above. Verify ROCm installation

/opt/rocm/bin/rocminfo /opt/rocm/opencl/bin/clinfo

If both two commands show the same GPU device, ROCm is successfully installed. Add ROCm binary path

echo 'export PATH=$PATH:/opt/rocm/bin:/opt/rocm/rocprofiler/bin:/opt/rocm/op

Remove ROCm packages

sudo apt autoremove rocm-opencl rocm-dkms rocm-dev rocm-utils rm -rf /opt/rocm* sudo reboot

Install dependency packages according to shawonashraf

sudo apt install rocm-libs hipcub miopen-hip

Ensure that you remove all rocm* packages in /opt folder

Install Tensorflow-ROCm https://medium.com/analytics-vidhya/install-tensorflow-2-for-amd-gpus-87e8d7aeb812

4/6

6/9/2021

Install Tensorflow 2 for AMD GPUs | by Eric Ngo | Analytics Vidhya | Medium

By typing

pip3 install tensorflow-rocm

Now, you may encounter errors of missing packages (e.g. hipsparse, rccl). To solve the errors, you should repeat the following commands to install missing packages as long as the errors disappear.

sudo apt install sudo ldconfig

Yup. That’s it. I tested the above commands on Ubuntu 18.04.5 (kernel 5.4) and Radeon VII. For CentosOS, follow this issue #1196. Hopefully, this post is helpful to you. If you have questions & comments for me, please leave comments or ping me. Have fun training Deep Learning models on AMD GPUs.

Sign up for Analytics Vidhya News Bytes https://medium.com/analytics-vidhya/install-tensorflow-2-for-amd-gpus-87e8d7aeb812

5/6

6/9/2021

Install Tensorflow 2 for AMD GPUs | by Eric Ngo | Analytics Vidhya | Medium

By Analytics Vidhya

Latest news from Analytics Vidhya on our Hackathons and some of our best articles!Take a look.

Your email Get this newsletter By signing up, you will create a Medium account if you don’t already have one. Review our Privacy Policy for more information about our privacy practices.

TensorFlow

Amd

Gpu

Rocm

Deep Learning

https://medium.com/analytics-vidhya/install-tensorflow-2-for-amd-gpus-87e8d7aeb812

6/6...