CNTK 2.2 released, Microsoft Deep Learning Toolkit

CNTK is the Microsoft Deep Learning Toolkit, first developed by Microsoft’s computer scientists who want to do their own research faster and more efficiently. It quickly surpassed the voice field and evolved into a product, including some of the leading international home appliance manufacturers and Microsoft’s flagship product groups (flagship product groups), including customers rely on it to perform a variety of deep learning tasks The It can easily handle from a relatively small to very very large and other large-scale data sets, either in a notebook running, you can also run in the data center in a series of computers. It can run on the use of traditional CPU or GPU on the computer; GPU previously used the main purpose is to deal with the graphics demanding video games, and later people can use it to be a very efficient operation of the depth of learning the required algorithm.

CNTK 2.2 has been released, this version adds some new features support and performance improvements, and the document was updated.

Changelog:

• This version requires cuDNN 6.0 to support extended convolution and deterministic pooling. Please update cuDNN.
• This version requires OpenCV to support the TensorBoard Image feature. Install OpenCV before installing CNTK.

download link:

Windows

• CNTK for Windows v.2.2 CPU only
• CNTK for Windows v.2.2 GPU
• CNTK for Windows v.2.2 GPU with 1bit-SGD
• NuGet Packages:
  • CNTK, CPU-Only Build
  • CNTK, GPU Build
  • CNTK, UWP CPU-Only Build
  • The CPU-only Model Evaluation CNTK the Libraries (MKL based)
    Package Penalty for the Description and Instructions are here Wallpaper.

Linux

• CNTK for Linux v.2.2 CPU only
• CNTK for Linux v.2.2 GPU
• CNTK for Linux v.2.2 GPU with 1bit-SGD
• Images at Docker Hub