# Features of DLAMI
<a name="features"></a>

The features of AWS Deep Learning AMIs (DLAMI) include preinstalled deep learning frameworks, GPU software, model servers, and model visualization tools.

## Preinstalled frameworks
<a name="features-frameworks"></a>

There are currently two primary flavors of DLAMI with other variations related to the operating system (OS) and software versions: 
+ [Deep Learning AMI with Conda](overview-conda.md) – Frameworks installed separately using `conda` packages and separate Python environments.
+ [Deep Learning Base AMI](overview-base.md) – No frameworks installed; only [NVIDIA CUDA](https://developer.nvidia.com/cuda-zone) and other dependencies.

The Deep Learning AMI with Conda uses `conda` environments to isolate each framework, so you can switch between them at will and not worry about their dependencies conflicting. The Deep Learning AMI with Conda supports the following frameworks:
+ PyTorch
+ TensorFlow 2

**Note**  
DLAMI no longer supports the following deep learning frameworks: Apache MXNet, Microsoft Cognitive Toolkit (CNTK), Caffe, Caffe2, Theano, Chainer, and Keras.

## Preinstalled GPU software
<a name="features-gpu"></a>

Even if you use a CPU-only instance, the DLAMIs will have [NVIDIA CUDA](https://developer.nvidia.com/cuda-zone) and [NVIDIA cuDNN](https://developer.nvidia.com/cudnn). The installed software is the same regardless of the instance type. Keep in mind that GPU-specific tools work only on an instance that has at least one GPU. For more information about instance types, see [Choosing a DLAMI instance type](instance-select.md).

For more information about CUDA, see [CUDA Installations and Framework Bindings](overview-cuda.md).

## Model serving and visualization
<a name="features-gpu"></a>

Deep Learning AMI with Conda comes preinstalled with model servers for TensorFlow, as well as TensorBoard for model visualizations. For more information, see [TensorFlow Serving](tutorial-tfserving.md).