Deep Learning AMI comes with popular deep learning frameworks optimized for high performance on Amazon EC2 instances. Includes Apache MXNet, TensorFlow, PyTorch, Caffe, Caffe2, Keras, Chainer, CNTK and Theano. The deep learning frameworks are installed in Conda environments to provide a reliable and isolated environment for machine learning practitioners. The Deep Learning AMI also includes Anaconda data science distribution, and is provided at no additional charge to Amazon EC2 users.
Deep Learning frameworks are pre-configured with latest versions of NVIDIA CUDA, cuDNN and Intel acceleration libraries such as MKL-DNN and iDeep for high performance across Amazon EC2 instances such as P3 and C5.

Deep Learning frameworks (with CUDA 8.0, 9.0, 9.1, 9.2 and cuDNN 7.3.1):
Apache MXNet 1.3.0. with Gluon
TensorFlow 1.12.0 + Horovod
Caffe2 0.8.1
PyTorch 0.4.1
Chainer 5.0.0
CNTK 2.6
Theano 1.0.3
Keras 2.2.4
Caffe 1.0 with CUDA 8 and cuDNN 6

Deep Learning AMI also comes with model debugging and hosting capabilities:
MXNet Model Server 1.0
TensorFlow Serving 1.12.0
TensorBoard 1.12.0
GPU Drivers: NVIDIA Driver 396.44
Get started with running your deep learning code in minutes:

Version

By

Operating System

Linux/Unix, Amazon Linux 2018.03

Delivery Methods

• Amazon Machine Image