Microsoft Cognitive Toolkit or CNTK is an open-source deep-learning toolkit. It is a state of the art production that is ready to use for deep learning.
CNTK is designed for ease of focusing on the problem when a user needs to solve training quicker on the large data sets.
CNTK makes Deep learning fast & scalable. It is used in a large number of production loads in the cloud environment. This Toolkit is tested in the production setting for accuracy, efficiency & scalability in the multi GPU, multi-server environment.
Several use cases of CNTK are speech, image, text & combination of data types deep learning applications. CNTK is used for object recognition in the fragmented reality application and in the face detection application.
It supports CUDA 10 for both Windows and Linux.
The latest version of CNTK is 2.7
The key training algorithms supported by CNTK are:
Convolutional Neural Networks (CNNs)
Recurrent Neural Networks (RNNs)
Long Short-Term Memory (LSTM)
CNTK Model can be trained by using programs like C++, C#/.NET, Python, and Java,
CNTK supports 64-bit Linux & Windows OS. You can either choose pre-compiled binary packages to install or compile the Toolkit from the source provided in GitHub.
With the help of Nvidia V100 GPU servers, customer can get dramatically faster training times and higher multi-node training performance with the latest versions of deep learning frameworks such as CNTK, Caffe2, MXNet, TensorFlow, etc
With the help of NGC (Nvidia GPU Cloud), which consists of deep learning framework containers, customers can use CNTK toolkit.
NVidia development tools also required to build the Microsoft CNTK and support libraries. After you install all the required NVidia tools, you should check that you have the latest graphic card driver installed.
Train your models on E2E GPU v100 Cloud servers here