Deep Learning AMI
Developer Guide

Release Note Details for Deep Learning AMI (Windows 2016) Version 1.0

AWS Deep Learning AMI

The AWS Deep Learning AMI are prebuilt with CUDA 8 and 9, and several deep learning frameworks.

Highlights of the Release

  1. CUDA 8 and 9

  2. CuDNN 6 and 7

  3. NCCL 2.0.5

  4. CuBLAS 8 and 9

  5. OpenCV 3.2.0

  6. SciPy 0.19.1

  7. Conda 5.01

Prebuilt Deep Learning Frameworks

  • Apache MXNet: MXNet is a flexible, efficient, portable and scalable open source library for deep learning. It supports declarative and imperative programming models, across a wide variety of programming languages, making it powerful yet simple to code deep learning applications. MXNet is efficient, inherently supporting automatic parallel scheduling of portions of source code that can be parallelized over a distributed environment. MXNet is also portable, using memory optimizations that allow it to run on mobile phones to full servers.

    • branch/tag used: v0.12.0

    • Built with CUDA 8 and cuDNN 6.1

    • Justification: Stable and well tested

    • Source Directory:

      C:\MXNet

  • Caffe: Caffe is a framework made with expression, speed, and modularity in mind.

    • branch/tag used: Windows branch, commit #5854

    • Built with CUDA 8 and cuDNN 5

    • Justification: Stable and well tested

    • Source Directory:

      C:\Caffe

  • TensorFlow: TensorFlow™ is an open source software library for numerical computation using data flow graphs.

    • branch/tag used : v1.4

    • Built with CUDA 8 and cuDNN 6.1

    • Justification : Stable and well tested

    • Source Directory:

      C:\ProgramData\Anaconda3\envs\tensorflow\Lib\site-packages

Python 2.7 and Python 3.5 Support

Python 2.7 and Python 3.6 are supported in the AMI for all of this installed Deep Learning Frameworks except Caffe2:

  1. Apache MXNet

  2. Caffe

  3. Tensorflow

Instance Type Support

P3 instance type is not yet supported.

The AMI supports all other CPU Instance Types for all frameworks.

GPU Drivers Installed

  • CuDNN 6 and 7

  • Nvidia 385.54

  • CUDA 8 and 9

Launching Deep Learning Instance

Choose the flavor of the AMI from the list below in the region of your choice and follow the steps at:

Launching a Windows Instance

Deep Learning AMI (Windows 2016) Regions

Windows 2016 AMIs are available in the following regions:

  • US East (Ohio): us-east-2 - ami-6cc1ef09

  • US East (N. Virginia): us-east-1 - ami-d50381af

  • GovCloud: us-gov-west-1 - ami-d5018db4

  • US West (N. California): us-west-1 - ami-0abf876a

  • US West (Oregon): us-west-2 - ami-d3be62ab

  • Beijing (China): cn-north-1 - ami-6f1e5000

  • Asia Pacific (Seoul): ap-northeast-2 - ami-9375d2fd

  • Asia Pacific (Singapore): ap-southeast-1 - ami-5c64323f

  • Asia Pacific (Sydney): ap-southeast-2 - ami-a0ca20c2

  • Asia Pacific (Tokyo): ap-northeast-1 - ami-c6fe42a0

  • Canada (Central): ca-central-1 - ami-4b82392f

  • EU (Frankfurt): eu-central-1 - ami-9e7efcf1

  • EU (Ireland): eu-west-1 - ami-8aee58f3

  • EU (London): eu-west-2 - ami-09edf26d

  • SA (Sao Paulo): sa-east-1 - ami-10bffa7c

References

Test Environments

  • Built on p2.16xlarge.

  • Also tested on p2.xlarge, c4.4xlarge.