Framework Support Policy

The framework support policy is live on the DLC dev guides.

Anaconda shifted to a commercial licensing model for certain users. Actively maintained DLCs have been migrated to the publicly available open-source version of Conda (conda-forge) from the Anaconda channel.

Warning: If you are actively using Anaconda to install and manage your packages and their dependencies in a DLC that is no longer actively maintained, you are responsible for complying with the governing license from the Anaconda Repository, if you determine that the terms apply to you. Alternatively, you can migrate to one of the currently-supported DLCs listed in the Supported Frameworks table or you can install packages using conda-forge as a source.

Available Deep Learning Containers Images

The following table lists the Docker image URLs that will be used by Amazon ECS in task definitions. Replace the <repository-name> and <image-tag> values based on your desired container.

Once you've selected your desired Deep Learning Containers image, continue with one of the following tutorials:

To run training and inference on Deep Learning Containers for Amazon EC2 using PyTorch and TensorFlow, see Amazon EC2 Tutorials
To run training and inference on Deep Learning Containers for Amazon ECS using PyTorch and TensorFlow, see Amazon ECS tutorials
To run training and inference on Hugging Face models in SageMaker, see Hugging Face on AWS
Deep Learning Containers for Amazon EKS offer CPU, GPU, and distributed GPU-based training, as well as CPU and GPU-based inference. To run training and inference on Deep Learning Containers for Amazon EKS using PyTorch, and TensorFlow, see Amazon EKS Tutorials
For information on security in Deep Learning Containers, see Security in AWS Deep Learning Containers
For a list of the latest Deep Learning Containers release notes, see Release Notes for Deep Learning Containers

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

Region	Code	General Container	Neuron Container	Example URL
US East (Ohio)	us-east-2	Available	Available	763104351884.dkr.ecr.us-east-2.amazonaws.com/<repository-name>:<image-tag>
US East (N. Virginia)	us-east-1	Available	Available	763104351884.dkr.ecr.us-east-1.amazonaws.com/<repository-name>:<image-tag>
US West (N. California)	us-west-1	Available	None	763104351884.dkr.ecr.us-west-1.amazonaws.com/<repository-name>:<image-tag>
US West (Oregon)	us-west-2	Available	Available	763104351884.dkr.ecr.us-west-2.amazonaws.com/<repository-name>:<image-tag>
Africa (Cape Town)	af-south-1	Available	None	626614931356.dkr.ecr.af-south-1.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Hong Kong)	ap-east-1	Available	None	871362719292.dkr.ecr.ap-east-1.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Hyderabad)	ap-south-2	Available	None	772153158452.dkr.ecr.ap-south-2.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Jakarta)	ap-southeast-3	Available	None	907027046896.dkr.ecr.ap-southeast-3.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Malaysia)	ap-southeast-5	Available	None	550225433462.dkr.ecr.ap-southeast-5.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Melbourne)	ap-southeast-4	Available	None	457447274322.dkr.ecr.ap-southeast-4.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Mumbai)	ap-south-1	Available	Available	763104351884.dkr.ecr.ap-south-1.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Osaka)	ap-northeast-3	Available	None	364406365360.dkr.ecr.ap-northeast-3.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Seoul)	ap-northeast-2	Available	None	763104351884.dkr.ecr.ap-northeast-2.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Singapore)	ap-southeast-1	Available	Available	763104351884.dkr.ecr.ap-southeast-1.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Sydney)	ap-southeast-2	Available	Available	763104351884.dkr.ecr.ap-southeast-2.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Taipei)	ap-east-2	Available	Available	763104351884.dkr.ecr.ap-east-2.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Thailand)	ap-southeast-7	Available	None	590183813437.dkr.ecr.ap-southeast-7.amazonaws.com/<repository-name>:<image-tag>
Asia Pacific (Tokyo)	ap-northeast-1	Available	Available	763104351884.dkr.ecr.ap-northeast-1.amazonaws.com/<repository-name>:<image-tag>
Canada (Central)	ca-central-1	Available	None	763104351884.dkr.ecr.ca-central-1.amazonaws.com/<repository-name>:<image-tag>
Canada (Calgary)	ca-west-1	Available	None	204538143572.dkr.ecr.ca-west-1.amazonaws.com/<repository-name>:<image-tag>
EU (Frankfurt)	eu-central-1	Available	Available	763104351884.dkr.ecr.eu-central-1.amazonaws.com/<repository-name>:<image-tag>
EU (Ireland)	eu-west-1	Available	Available	763104351884.dkr.ecr.eu-west-1.amazonaws.com/<repository-name>:<image-tag>
EU (London)	eu-west-2	Available	None	763104351884.dkr.ecr.eu-west-2.amazonaws.com/<repository-name>:<image-tag>
EU (Milan)	eu-south-1	Available	None	692866216735.dkr.ecr.eu-south-1.amazonaws.com/<repository-name>:<image-tag>
EU (Paris)	eu-west-3	Available	Available	763104351884.dkr.ecr.eu-west-3.amazonaws.com/<repository-name>:<image-tag>
EU (Spain)	eu-south-2	Available	None	503227376785.dkr.ecr.eu-south-2.amazonaws.com/<repository-name>:<image-tag>
EU (Stockholm)	eu-north-1	Available	None	763104351884.dkr.ecr.eu-north-1.amazonaws.com/<repository-name>:<image-tag>
EU (Zurich)	eu-central-2	Available	None	380420809688.dkr.ecr.eu-central-2.amazonaws.com/<repository-name>:<image-tag>
Israel (Tel Aviv)	il-central-1	Available	None	780543022126.dkr.ecr.il-central-1.amazonaws.com/<repository-name>:<image-tag>
Mexico (Central)	mx-central-1	Available	None	637423239942.dkr.ecr.mx-central-1.amazonaws.com/<repository-name>:<image-tag>
Middle East (Bahrain)	me-south-1	Available	None	217643126080.dkr.ecr.me-south-1.amazonaws.com/<repository-name>:<image-tag>
Middle East (UAE)	me-central-1	Available	None	914824155844.dkr.ecr.me-central-1.amazonaws.com/<repository-name>:<image-tag>
South America (Sao Paulo)	sa-east-1	Available	Available	763104351884.dkr.ecr.sa-east-1.amazonaws.com/<repository-name>:<image-tag>
China (Beijing)	cn-north-1	Available	None	727897471807.dkr.ecr.cn-north-1.amazonaws.com.cn/<repository-name>:<image-tag>
China (Ningxia)	cn-northwest-1	Available	None	727897471807.dkr.ecr.cn-northwest-1.amazonaws.com.cn/<repository-name>:<image-tag>

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:2.9.1-gpu-py39-cu112-ubuntu20.04-ec2

Important

You must login to access the DLC image repository before pulling the image. Ensure your CLI is up to date using the steps in Installing the current AWS CLI Version Then, specify your region and its corresponding ECR Registry from the previous table in the following command:

aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 763104351884.dkr.ecr.us-east-1.amazonaws.com

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

docker pull <name of container image>

DLC Available Image User Guide

To use the following tables, select your desired framework, the kind of job you're starting, and your desired Python version. Your job type is either training, inference or general. Your Python version is either py37, py38, py39, py310, py311 or py312 depending on availability. Plug this information into the replaceable portions of the URL as shown in the example URL.

You can pin your version by adding the version tag to your URL as follows:

763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.4.1-cpu-py37-ubuntu18.04-v1.0

Base Containers

Available in Amazon ECR Public Gallery: https://gallery.ecr.aws/deep-learning-containers/base

vLLM Containers

Available in Amazon ECR Public Gallery: https://gallery.ecr.aws/deep-learning-containers/vllm

EC2 Framework Containers (EC2, ECS, and EKS support only)

Framework	Job Type	Horovod Options	CPU/GPU	Python Version Options	Example URL
PyTorch 2.8.0	training	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.8.0-cpu-py312-ubuntu22.04-ec2
PyTorch 2.8.0	training	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.8.0-gpu-py312-cu129-ubuntu22.04-ec2
PyTorch 2.6.0	inference	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference:2.6.0-cpu-py312-ubuntu22.04-ec2
PyTorch 2.6.0	inference	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference:2.6.0-gpu-py312-cu124-ubuntu22.04-ec2
TensorFlow 2.18.0	training	No	CPU	3.10 (py310)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.18.0-cpu-py310-ubuntu22.04-ec2
TensorFlow 2.18.0	training	No	GPU	3.10 (py310)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.18.0-gpu-py310-cu125-ubuntu22.04-ec2
TensorFlow 2.18.0	inference	No	CPU	3.10 (py310)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference:2.18.0-cpu-py310-ubuntu20.04-ec2
TensorFlow 2.18.0	inference	No	GPU	3.10 (py310)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference:2.18.0-gpu-py310-cu122-ubuntu20.04-ec2

SageMaker Framework Containers (SM support only)

Framework	Job Type	Horovod Options	CPU/GPU	Python Version Options	Example URL
PyTorch 2.8.0	training	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.8.0-cpu-py312-ubuntu22.04-sagemaker
PyTorch 2.8.0	training	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.8.0-gpu-py312-cu129-ubuntu22.04-sagemaker
PyTorch 2.6.0	inference	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference:2.6.0-cpu-py312-ubuntu22.04-sagemaker
PyTorch 2.6.0	inference	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference:2.6.0-gpu-py312-cu124-ubuntu22.04-sagemaker
TensorFlow 2.19.0	training	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.19.0-cpu-py312-ubuntu22.04-sagemaker
TensorFlow 2.19.0	training	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-training:2.19.0-gpu-py312-cu125-ubuntu22.04-sagemaker
TensorFlow 2.19.0	inference	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference:2.19.0-cpu-py312-ubuntu22.04-sagemaker
TensorFlow 2.19.0	inference	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference:2.19.0-gpu-py312-cu122-ubuntu22.04-sagemaker

EC2 Framework ARM64/Graviton Containers (EC2, ECS, and EKS support only)

Important note: Starting with PyTorch 2.5, we are changing the name of Graviton DLCs to ARM64 DLCs in order to generalize the usage. For example, the ECR repository name is now "pytorch-inference-arm64" instead of "pytorch-inference-graviton". Graviton DLCs and ARM64 DLCs are functionally equivalent.

Framework	Job Type	Horovod Options	CPU/GPU	Python Version Options	Example URL
PyTorch 2.7.0	training	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training-arm64:2.7.0-gpu-py312-cu128-ubuntu22.04-ec2
PyTorch 2.6.0	inference	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference-arm64:2.6.0-cpu-py312-ubuntu22.04-ec2
PyTorch 2.6.0	inference	No	GPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference-arm64:2.6.0-gpu-py312-cu124-ubuntu22.04-ec2
TensorFlow 2.18.0	inference	No	CPU	3.10 (py310)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference-arm64:2.18.0-cpu-py310-ubuntu20.04-ec2

SageMaker Framework ARM64/Graviton Containers (SM support only)

Important note: Starting with PyTorch 2.5, we are changing the name of Graviton DLCs to ARM64 DLCs in order to generalize the usage. For example, the ECR repository name is now "pytorch-inference-arm64" instead of "pytorch-inference-graviton". Graviton DLCs and ARM64 DLCs are functionally equivalent.

Framework	Job Type	Horovod Options	CPU/GPU	Python Version Options	Example URL
PyTorch 2.6.0	inference	No	CPU	3.12 (py312)	763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-inference-arm64:2.6.0-cpu-py312-ubuntu22.04-sagemaker
TensorFlow 2.18.0	inference	No	CPU	3.10 (py310)	763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference-arm64:2.18.0-cpu-py310-ubuntu20.04-sagemaker

NVIDIA Triton Inference Containers (SM support only)

Versions 23.12 and onwards:

Starting version 23.12, onwards, the sagemaker tritonserver is available in a different set of accounts than the previous ones. The new accounts are now the same as other DLCs listed on this page at the top, making it easier to switch containers going forward. These accounts are listed in section '# Available Deep Learning Containers Images' above.
They can now be obtained programmatically from the sagemaker python sdk as:

from sagemaker import image_uris

triton_framework = "sagemaker-tritonserver"
region="us-west-2"
instance_type="ml.g5.12xlarge"

available_versions = list(image_uris.config_for_framework(triton_framework)['versions'].keys())
image_uri = image_uris.retrieve(framework=triton_framework, region=region, instance_type=instance_type, version=available_versions[0])

Available versions: 25.09, 25.04, 24.09, 24.05, 24.03, 24.01, 23.12

The Sagemaker Triton inference containers are built on top of the NGC containers with SageMaker support. To identify the python version and versions for other packages please refer to the corresponding official release notes for the specific version here: https://docs.nvidia.com/deeplearning/triton-inference-server/release-notes/index.html

Versions prior to 23.12:

For versions prior to 23.12, the following 23.<XY> versions are available: 23.01, 23.02, 23.03, 23.05, 23.06, 23.07, 23.08, 23.09, 23.10.
For versions of the 22.<XY> series, the following are available: 22.05, 22.07, 22.08, 22.09, 22.10, 22.12
For versions of the 21.<XY> series, the following are available: 21.08
The following example notebook demonstrates the account_id_map to obtain the account for versions prior to r23.12: https://github.com/aws/amazon-sagemaker-examples/blob/main/sagemaker-triton/resnet50/triton_resnet50.ipynb

Framework	Job Type	Horovod Options	CPU/GPU	Python Version Options	Example URL
NVIDIA Triton Inference Server 23.`<XY>`	inference	No	GPU	3.8 (py38)	007439368137.dkr.ecr.us-east-2.amazonaws.com/sagemaker-tritonserver:23.`<XY>`-py3
NVIDIA Triton Inference Server 23.`<XY>`	inference	No	CPU	3.8 (py38)	007439368137.dkr.ecr.us-east-2.amazonaws.com/sagemaker-tritonserver:23.`<XY>`-py3-cpu

Note:

SageMaker Triton Inference Container does not support Tensorflow1 as of version 23.05 onwards, as upstream Triton container does not support Tensorflow(v1) native backend from version 23.04 onwards.
SageMaker Triton Inference Container does not ship with the FasterTransformer(FT) backend from version 23.06 onwards since the upstream FT library is undergoing re-structuring. It was previously available from versions v22.12 - v23.05, experimentally.

Large Model Inference Containers

Starting LMI V10 (0.28.0), we are changing the name from LMI DeepSpeed DLC to LMI (LargeModelInference). As part of this change, we have decided to discontinue integration with DeepSpeed library into the container. You can continue to use vLLM or LMI-dist Library with the LMI container. If you plan to use DeepSpeed Library, please follow the steps here or use LMI V9 (0.27.0).

Framework	Job Type	Accelerator	Python Version Options	Example URL
DJLServing 0.34.0 with vLLM 0.10.2, Transformers 4.55.2, and Accelerate 1.0.1	inference	GPU	3.12 (py312)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.34.0-lmi16.0.0-cu128
DJLServing 0.33.0 with vLLM 0.8.4, Transformers 4.51.3, and Accelerate 1.0.1	inference	GPU	3.12 (py312)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.33.0-lmi15.0.0-cu128
DJLServing 0.33.0 with TensorRT-LLM 0.21.0rc1, Transformers 4.51.3, and Accelerate 1.0.1	inference	GPU	3.12 (py312)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.33.0-tensorrtllm0.21.0-cu128
DJLServing 0.32.0 with LMI Dist 13.0.0, vLLM 0.7.1, Transformers 4.45.2, and Accelerate 1.0.1	inference	GPU	3.11 (py311)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.32.0-lmi14.0.0-cu126
DJLServing 0.32.0 with TensorRT-LLM 0.12.0, Transformers 4.44.2, and Accelerate 0.32.1	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.32.0-tensorrtllm0.12.0-cu125
DJLServing 0.31.0 with LMI Dist 13.0.0, vLLM 0.6.3.post1, Transformers 4.45.2, and Accelerate 1.0.1	inference	GPU	3.11 (py311)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.31.0-lmi13.0.0-cu124
DJLServing 0.30.0 with LMI Dist 12.0.0, vLLM 0.6.2, Transformers 4.45.2, and Accelerate 1.0.1	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.30.0-lmi12.0.0-cu124
DJLServing 0.30.0 with TensorRT-LLM 0.12.0, Transformers 4.44.2, and Accelerate 0.33.0	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.30.0-tensorrtllm0.12.0-cu125
DJLServing 0.30.0 with Neuron SDK 2.20.1, TransformersNeuronX 0.12.313, and Transformers 4.45.2	inference	Neuron	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.30.0-neuronx-sdk2.20.1
DJLServing 0.29.0 with TensorRT-LLM 0.11.0, Transformers 4.42.4, and Accelerate 0.32.1	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.29.0-tensorrtllm0.11.0-cu124
DJLServing 0.29.0 with LMI Dist 11.0.0, Transformers 4.43.2, Accelerate 0.32.1	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.29.0-lmi11.0.0-cu124
DJLServing 0.29.0 with Neuron SDK 2.19.1, TransformersNeuronX 0.11.351 and Transformers 4.43.1	inference	Neuron	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.29.0-neuronx-sdk2.19.1
DJLServing 0.28.0 with TensorRT-LLM 0.9.0, Transformers 4.40.0, and Accelerate 0.29.3	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.28.0-tensorrtllm0.9.0-cu122
DJLServing 0.28.0 with LMI Dist 0.10.0, Transformers 4.41.1, Accelerate 0.30.1	inference	GPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.28.0-lmi10.0.0-cu124
DJLServing 0.28.0 with Neuron SDK 2.18.2, TransformersNeuronX 0.10.0.360 and Transformers 4.36.2	inference	Neuron	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.28.0-neuronx-sdk2.18.2

DJL CPU Full Inference Containers

Framework	Job Type	CPU/GPU	Python Version Options	Example URL
DJLServing 0.29.0	inference	CPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.29.0-cpu-full
DJLServing 0.28.0	inference	CPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.28.0-cpu-full
DJLServing 0.27.0	inference	CPU	3.10 (py310)	763104351884.dkr.ecr.us-west-2.amazonaws.com/djl-inference:0.27.0-cpu-full

AutoGluon Training Containers

Framework	AutoGluon Version	Job Type	CPU/GPU	Python Version Options	Example URL
AutoGluon 1.4.0	1.4.0	training	GPU	3.11 (py311)	763104351884.dkr.ecr.us-west-2.amazonaws.com/autogluon-training:1.4.0-gpu-py311-cu124-ubuntu22.04
AutoGluon 1.4.0	1.4.0	training	CPU	3.11 (py311)	763104351884.dkr.ecr.us-west-2.amazonaws.com/autogluon-training:1.4.0-cpu-py311-ubuntu22.04

AutoGluon Inference Containers

Framework	AutoGluon Version	Job Type	CPU/GPU	Python Version Options	Example URL
AutoGluon 1.4.0	1.4.0	inference	GPU	3.11 (py311)	763104351884.dkr.ecr.us-west-2.amazonaws.com/autogluon-inference:1.4.0-gpu-py311-cu124-ubuntu22.04
AutoGluon 1.4.0	1.4.0	inference	CPU	3.11 (py311)	763104351884.dkr.ecr.us-west-2.amazonaws.com/autogluon-inference:1.4.0-cpu-py311-ubuntu22.04

HuggingFace Training Containers

Please refer to the following page to view all available versions and tags for GPU containers:

GPU Release Page