Deep Learning Containers Images - AWS Deep Learning Containers

Deep Learning Containers Images

AWS Deep Learning Containers are available as Docker images in Amazon ECR. Each Docker image is built for training or inference on a specific Deep Learning framework version, python version, with CPU or GPU support. The following table lists the Docker image URLs that will be used by Amazon ECS in task definitions.

Deep Learning Containers Docker Images are available in the following regions:

Region

Code

General Container

Elastic Inference Container

US East (N. Virginia)

us-east-1

Available

Available

US East (Ohio)

us-east-2

Available

Available

US West (N. California)

us-west-1

Available

None

US West (Oregon)

us-west-2

Available

Available

Asia Pacific (Tokyo)

ap-northeast-1

Available

Available

Asia Pacific (Seoul)

ap-northeast-2

Available

Available

Asia Pacific (Hong Kong)

ap-east-1

Available

None

Asia Pacific (Mumbai)

ap-south-1

Available

None

Asia Pacific (Singapore)

ap-southeast-1

Available

None

Asia Pacific (Sydney)

ap-southeast-2

Available

None

Middle East (Bahrain)

me-south-1

Available

None

South America (Sao Paulo)

sa-east-1

Available

None

Canada (Central)

ca-central-1

Available

None

EU (Frankfurt)

eu-central-1

Available

None

EU (Stockholm)

eu-north-1

Available

None

EU (Ireland)

eu-west-1

Available

Available

EU (London)

eu-west-2

Available

None

EU (Paris)

eu-west-3

Available

None

Note: eu-north-1 is available starting in the version 2.0 release. As a result, MXNet-1.4.0 is not available in this region.

ECR is a regional service and the Image table contains the URLs for us-east-1 images. To pull from one of the regions mentioned previously, insert the region in the repository URL following this example:

763104351884.dkr.ecr.<region>.amazonaws.com/tensorflow-training:1.15.2-cpu-py27-ubuntu18.04
Important

You must login to access to the Deep Learning Containers image repository before pulling the image. Specify your region in the following command:

$(aws ecr get-login --no-include-email --region us-east-1 --registry-ids 763104351884)

You can then pull these Docker images from Amazon ECR by running:

docker pull <name of container image>

General Framework Containers

To use the following table, select your desired framework, as well as the kind of job you're starting, and the desired Python version. Your job type is either training or inference. Your Python version is either py27 or py36. Plug this information into the replaceable portions of the URL as shown in the example URL.

Note

If you use the ap-east-1 region, replace the account ID "763104351884" with "871362719292". If you use the me-south-1 region, replace the account ID "763104351884" with "217643126080".

Framework

Job Type

Horovod Options

CPU/GPU

Python Version Options

Example URL

TensorFlow 2.1

training, inference

Training:Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.1.0-cpu-py27-ubuntu18.04

TensorFlow 2.1

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.1.0-gpu-py27-cu101-ubuntu18.04

TensorFlow 1.15.2

training, inference

Training:Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:1.15.2-cpu-py27-ubuntu18.04

TensorFlow 1.15.2

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:1.15.2-gpu-py27-cu100-ubuntu18.04

MXNet 1.6.0

training, inference

Training: Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/mxnet-training:1.6.0-cpu-py27-ubuntu16.04

MXNet 1.6.0

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/mxnet-training:1.6.0-gpu-py27-cu101-ubuntu16.04
PyTorch 1.4.0

training

No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.4.0-cpu-py27-ubuntu16.04
PyTorch 1.4.0

inference

No

CPU

3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference:1.4.0-cpu-py36-ubuntu16.04
PyTorch 1.4.0

training

Yes

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.4.0-gpu-py27-cu101-ubuntu16.04
PyTorch 1.4.0

inference

No

GPU

3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference:1.4.0-gpu-py36-cu101-ubuntu16.04

Elastic Inference Containers

Framework

Job Type

Horovod Options

CPU/GPU

Python Version Options

Example URL

TensorFlow 1.14.0 with Elastic Inference

inference

No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference-eia:1.14.0-cpu-py27-ubuntu16.04

MXNet 1.4.1 with Elastic Inference

inference

No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/mxnet-inference-eia:1.4.1-cpu-py27-ubuntu16.04

PyTorch 1.3.1 with Elastic Inference

Inference

No

CPU

3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference-eia:1.3.1-cpu-py36-ubuntu16.04

Prior Versions

Framework

Job Type

Horovod Options

CPU/GPU

Python Version Options

Example URL

TensorFlow 2.0.1

training, inference

Training:Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.0.1-cpu-py27-ubuntu18.04

TensorFlow 2.0.1

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.0.1-gpu-py27-cu100-ubuntu18.04

TensorFlow 2.0.0

training, inference

Training:Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.0.0-cpu-py27-ubuntu18.04

TensorFlow 2.0.0

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.0.0-gpu-py27-cu100-ubuntu18.04

TensorFlow 1.15.0

training, inference Training:Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:1.15.0-cpu-py27-ubuntu18.04

TensorFlow 1.15.0

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:1.15.0-gpu-py27-cu100-ubuntu18.04

TensorFlow 1.14.0

training, inference

Training:Yes, Inference: No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:1.14.0-cpu-py27-ubuntu16.04

TensorFlow 1.14.0

training, inference

Training: Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:1.14.0-gpu-py27-cu100-ubuntu16.04

MXNet 1.4.1

training, inference

No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/mxnet-training:1.4.1-cpu-py27-ubuntu16.04

MXNet 1.4.1

training, inference

No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/mxnet-training:1.4.1-gpu-py27-cu100-ubuntu16.04
PyTorch 1.3.1

training, inference

No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.3.1-cpu-py27-ubuntu16.04
PyTorch 1.3.1

training, inference

Training:Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.3.1-gpu-py27-cu101-ubuntu16.04
PyTorch 1.2.0

training, inference

No

CPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.2.0-cpu-py27-ubuntu16.04
PyTorch 1.2.0

training, inference

Training:Yes, Inference: No

GPU

2.7 (py27), 3.6 (py36)

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.2.0-gpu-py27-cu100-ubuntu16.04