What is Cuda Deep Learning

CUDA installations and framework bindings

Overall, deep learning is very new. However, every framework provides "stable" versions. These stable releases may not work with the latest CUDA or cuDNN implementations and features. Which variant should you choose? This ultimately depends on your use case and the functions you need. If in doubt, take the latest deep learning AMI with Conda. It has official Pip binaries of all frameworks with CUDA 10, depending on which latest version is supported by the particular framework. If you want the latest versions and want to customize your deep learning environment, choose Deep Learning Base AMI.

For more information, see our guide in Stable versus Release Candidates.

Selecting a DLAMI with CUDA

The AWS Deep Learning Base has CUDA 10, 10.1 and 10.2.

The Deep Learning AMI with Conda has CUDA 10, 10.1 and 10.2.

  • CUDA 10.1 with cuDNN 7: Apache MxNet (incubate), PyTorch and TensorFlow 2

  • CUDA 10 with cuDNN 7: TensorFlow and Chainer

We no longer include the CNTK, Caffe, Caffe2 and Theano Conda environments in the AWS Deep Learning AMI from version v28. Earlier versions of the AWS Deep Learning AMI that contain these environments are still available. However, we will only provide updates for these environments when security fixes for these frameworks are released by the open source community.

For information about installation options for DLAMI types and operating systems, see the pages for each CUDA version and option:

For specific framework version numbers, see the DLAMI Release Notes

Select this DLAMI type, or use the suggestion under “Next Topic” to learn more about the different DLAMIs.

Select one of the CUDA versions and see the complete list of DLAMIs with this version in the appendix. For more information on the various DLAMIs, see the "Next Topic" link.

next topic

DLAMI operating system options are available.

Related topics