-
Notifications
You must be signed in to change notification settings - Fork 4.3k
Home
This page has migrated to our new site. Please update any bookmarks.
The Microsoft Cognitive Toolkit - CNTK - is a unified deep-learning toolkit by Microsoft. This video provides a high-level view of the toolkit.
It can be included as a library in your Python or C++ programs, or used as a standalone machine learning tool through its own model description language (BrainScript). CNTK supports 64-bit Linux or 64-bit Windows operating systems. To install you can either choose pre-compiled binary packages, or compile the Toolkit from the source provided in Github.
The latest release of the Microsoft Cognitive Toolkit is v 2.0, released on June 1st, 2017.
Here are a few pages to get started:
- Eight reasons to switch from TensorFlow to CNTK
- Setting up CNTK on your machine
-
Tutorials, Examples, etc..
- Try the tutorials on Azure Notebooks with pre-installed CNTK
- The CNTK Library APIs
- CNTK as a machine learning tool through BrainScript
- How to contribute to CNTK
- Give us feedback through these channels.
2017-06-01. CNTK 2.0 The first production release of Cognitive Toolkit v.2.
Highlights:
- CNTK backend for Keras.
- Extremely fast binary convolution with Halide.
- Java API.
- A new set of NuGet Packages.
- Multiple bug fixes.
See more in the Release Notes. Get the Release from the CNTK Releases page.
2017-05-24. CNTK 2.0 Release Candidate 3
Release Candidate 3 is the final preview of Cognitive Toolkit v.2.0.
Highlights:
- API that were previously declared deprecated are now removed. See details in release notes.
- Introduction of CNTK Java API in experimental mode. See details in release notes.
- New operators like
to_sequence
andsequence.unpack
. - Support of convolution in 1D.
- Support of UDF serialization (available both in Python and native in C++).
- New tools (Crosstalk and RNN Conversion).
- Support of NVIDIA cuDNN v.6.0 when CNTK is built by the user from source code.
- A new set of NuGet Packages.
- Multiple bug fixes.
See more in the Release Notes.
Get the Release from the CNTK Releases page.
2017-04-21. CNTK 2.0 Release Candidate 2
With Release Candidate 2 we reacted to customer feedback and improved/added features, functionality, and performance.
Highlights:
- New operators like
pow
,sequence.reduce_max
,sequence.softmax
. - New feature for Linux source builds (GPU Direct RDMA support in distributed gradients aggregation, NCCL support for Python in V2 gradients aggregation).
- Support for Python 3.6 for source and binary installation; see here.
-
UserMinibatchSource
to write custom minibatch sources; see here. - New C# APIs:
class NDArrayView
and methods,SetMaxNumCPUThreads()
,GetMaxNumCPUThreads()
,SetTraceLevel()
,GetTraceLevel()
- A new set of NuGet Packages is provided with this Release.
The release notes contain an overview. Get the release from the CNTK Releases Page.
2017-03-31. CNTK 2.0 Release Candidate 1 With Release Candidate 1 the Microsoft Cognitive Toolkit enters the final set of enhancements before release of the production version of CNTK 2.0.
Highlights:
- The release candidate contains all changes and improvements introduced in CNTK 2.0 during beta phase.
- Enables Caffe-converted pretrained models on image classification including AlexNet, ResNet, VGG and BN-Inception.
- Slice now supports multiple-axis slicing.
- Improves performance and memory footprint
- Improvements in the device selection API.
- New Python model debugging functions.
- Improvements in Python and C# API. See the release notes for detailed description.
- New file names for CNTK libraries and dlls.
2017-03-16. V 2.0 Beta 15 Release available at Docker Hub
CNTK V 2.0 Beta 15 Runtime packages are now available as Public Images at Docker Hub.
See more on CNTK as Docker Images in this Wiki article.
2017-03-15. V 2.0 Beta 15 Release
Highlights of this Release:
- In addition to pre-existing python support, added support for TensorBoard output in BrainScript. Read more here.
- Learners can now be implemented in pure Python by means of
UserLearners
. Read more here. - New debugging helpers:
dump_function()
,dump_signature()
. - Tensors can be indexed using advanced indexing. E.g.
x[[0,2,3]]
would return a tensor that contains the first, third and fourth element of the first axis. - Significant updates in the Layers Library of Python API. See Release Notes for detailed description.
- Updates and new examples in C# API.
- Various bug fixes.
See more in the Release Notes.
Get the Release from the CNTK Releases page.
See all news.