Posted On: Oct 15, 2018
Faster training with optimized TensorFlow 1.11
The Deep Learning AMIs now come with an optimized build of TensorFlow 1.11, custom-built from source to accelerate deep learning applications on Amazon EC2 C5 and P3 instances. For Amazon EC2 C5 instances, Deep Learning AMIs deploy compute-optimized TensorFlow built with Intel Advanced Vector Extensions (AVX instruction sets) to speed up the performance of vector and floating point operations. The AMIs also come pre-configured to leverage Intel Math Kernel Library for Deep Neural Networks (MKL-DNN). Training a ResNet-50 benchmark with the synthetic ImageNet dataset using our optimized build of TensorFlow 1.11 on a c5.18xlarge instance type was 11X faster than training on the stock TensorFlow 1.11 binaries.
For Amazon EC2 P3 instances, Deep Learning AMIs deploy a TensorFlow build pre-configured with latest NVIDIA CUDA 9.0 and cuDNN 7.3.1 for leveraging the mixed precision floating point computation capabilities of the Volta V100 GPUs. Deep Learning AMIs automatically deploy the framework builds optimized for the EC2 instance you select when you activate the framework's virtual environment for the first time.
AWS Deep Learning AMIs also support popular frameworks including PyTorch, and Apache MXNet —all pre-installed and fully-configured for you to start developing your deep learning models in minutes while taking full advantage of the computational power of Amazon EC2 instances. For a complete list of frameworks and versions supported by the AWS Deep Learning AMI, see the release notes.
Get started quickly with the AWS Deep Learning AMIs using the getting-started guides and beginner-to-advanced level tutorials in our developer guide. You can also subscribe to our discussion forum to get launch announcements and post your questions.