Supported Tags and respective Dockerfile links

• latest ( latest/Dockerfile ) ⁠

Based On

https://github.com/avmgithub/pytorch_builder.git⁠

Usage Information

Dockerfile README

The Dockerfile creates a container that builds and installs pytorch in a pytorch user environment.
The base docker container is from https://hub.docker.com/r/nvidia/cuda-ppc64le (tag 9.0-cudnn7-devel-ubuntu16.04)
The python (version 3.6) used is from Anaconda and is installed in /home/pytorch/miniconda
The main script it executes is cloned from https://github.com/avmgithub/pytorch_builder.git⁠
The build_nimbix.sh also builds magma-cuda 2.3.0 from http://icl.utk.edu/projectsfiles/magma/downloads/magma-2.3.0.tar.gz⁠

There are 4 ENV variables that default to certain values. Feel free to change them as desired.

• ENV username=jenkins
• ENV python_version=3
• ENV git_commit=HEAD
• ENV branch=master

Instructions to build container

git clone https://github.com/ppc64le/build-scripts.git⁠
cd build-scripts/pytorch/Dockerfiles/latest_ubuntu_16.04
sudo docker build --tag="pytorch" .

Note

This supports CPU-only pytorch builds

Publication Statement

"This content is the sole responsibility of the entity that makes it available. IBM may review content to determine whether it is illegal or violates IBM policies, and IBM may remove or refuse to display content that IBM reasonably believe violates IBM policies or the law. For more information please see:

DMCA Takedown Policy: https:https://help.github.com/articles/dmca-takedown-policy⁠ IBM Takedown and DMCA Notices Website: https://www.ibm.com/legal/us/en/dmca.html⁠ IBM Copyright and Trademark Information Website: https://www.ibm.com/legal/us/en/copytrade.shtml⁠."

Please contact ich@us.ibm.com⁠ ( if this content is found to be in violation of the DMCA or any related licenses or publishing conditions.

Community Support

For any questions on the publication of this container, please contact us at aschuh@illinois.edu⁠.