RocketCE Pytorch includes support for IBM's Distributed Deep Learning (DDL).
This version of Pytorch will have the support of matrix-multiply assist (MMA) which was introduced in Power 10.
This release of RocketCE is based on Release OpenCE v1.3.1 and it includes Pytorch 1.8.1.
These packages are supported for both Power 9 and Power 10 Systems of IBM with linux-ppc64le.
The CPU-only variant of Pytorch is built without GPU and CUDA support. It has a smaller installation size, and omits features that would require a GPU. It does not include support for DDL, LMS, or NVIDIA's Apex.
This version of Pytorch will not support GPU variant. Support of GPU variant will be provided in coming releases. For more information, refer main page of OpenCE
To install pytorch built for CPU support run the following command:
Once the installation is success, validate the installation by running the following command which lists the package pytorch which was pulled from rocketce channel
Pytorch conda packages
Pytorch package from RocketCE also includes the following conda packages
The package-base packages come with CPU variants, and include CPU in the build string.
More informationThis conda package is using the Power 10 MMA capabilities. More information is available in Matrix-multiply assist (MMA) capabilities.
This version of Pytorch package will support ppc64le architecture on Power machines. Power 10 architecture has enhancements for the Matrix-Multiply Assist (MMA) capability, and these packages are build using openblas package which will utilize the MMA capability by which we can see better performance than regular packages.
77 4th AvenueWaltham, MA 02451 USA
Rocket Support Community
All Support Offerings
About Rocket Software
Training and Services
Forum Terms and Conditions
Contact Forum Moderator