AWS Deep Learning AMIs now comes with optimized TensorFlow 1.8 to deliver training on P3 instances and EC2 C5 instances

The AWS Deep Learning AMIs for Amazon Linux and Ubunty now comes with the optimization for TensorFlow 1.8 to offer higher-performance training for Amazon Elastic Cloud Compute C5 instances and P3 instances. TensorFlow 1.8 is built with the Intel’s advanced vector instructions, FMA and SSE instruction sets to increase vector and floating point computations. The AMIs is fully configured with the Intel MKL-DNN to speed up the math routines utilized in neural network training on Amazon Elastic Cloud Compute C5 instances. By enhancing the training performance for GPU based scenarios the AMIs consist of an optimized build of TensorFlow 1.8 fully configured with cuDNN 7 and NVIDIA CUDA 9 to take advantage of the mixed precision training on Volta V100 GPUs accelerating Amazon Elastic Cloud Compute P3 instances.