DeepLearning Images Revision M9. Intel optimized images!
With this release we have done a huge step forward in improving user experience of the users on our images.
Release has been done with partnership with Intel in order to provide the fastest possible set of ML/DL frameworks. This is the first (according to my knowledge) available on the market VM image that not only includes MKL optimized TensorFlow binaries but also has Intel optimized main popular ML framework:
- Intel® Optimized numpy
- Intel® Optimized scipy
- Intel® Optimized scikit-learn
You can learn more about the Intel packages here. Here is a benchmarks (of the numpy) result from the official web site:
Since these are generic packages they are pre-installed on ALL images. Even if you are using GPU images your pre/post processing work that is based on numpy/scipy/scikit-learn will benefit from these packages as well!
There are no new flavors introduced with these release. As usual you can create Deep Learning VM via either Google Marketplace/Google Compute Engine or CLI.
