This project will no longer be maintained by Intel. Intel has ceased development and contributions including, but not limited to, maintenance, bug fixes, new releases, or updates, to this project. Intel no longer accepts patches to this project.
This repo contains code and instructions for launching a pre-configured Azure Data Science Virtual Machine (DSVM) for Linux with CPU-optimized TensorFlow, MXNet and PyTorch. For more information, see our recent blog detailing some optimizations and startup instructions.
The Intel® Optimized Data Science Virtual Machine is an extension of the Ubuntu version of Microsoft's DSVM and comes with Python environments optimized for deep learning on Intel® Xeon® Processors. These environments include open source deep learning frameworks with Intel® MKL-DNN as a backend for optimal performance on Intel® Xeon Processors. These environments require no changes to existing code and accelerate deep learning training and inference. Additionally, this offering includes all software packages available on the base DSVM with several popular tools for data science and ML which are already pre-installed, configured and tested. For more info on the Azure DSVM, click here. For additional information on Intel® Optimizations for deep learning frameworks, please click here.
- Compute Optimized: Fsv2-series (F4sv2, F8sv2, F16sv2, F32sv2, F64sv2, F72sv2)
- High Performance Compute: Hc-series
Click here to view the offer in the Azure Marketplace, and click GET IT NOW and follow the prompts to launch the VM. See this post for step by step instructions.
Note: This VM takes about 10 minutes to launch. At creation, a custom extension triggers a one-time installation of the latest Intel® Optimized deep learning frameworks. Once launched, you will be able to start and stop the VM as usual. Instructions are organized into two sections:
- Display available virtual environments with
conda env list - Activate the desired virtual environment with
source activate <env_name>(ex:source activate intel_tensorflow_p3) - To run benchmarks, follow instructions here