ML.NET 1.0 RC released. What's new?

Original author: Cesar De la Torre
  • Transfer
ML.NET is an open source, cross-platform machine learning environment (Windows, Linux, macOS) for .NET developers. Working with ML.NET, developers can use existing tools and skills to develop and implement AI in their applications, creating custom machine learning models for common scenarios such as Sentiment Analysis, Recommendation, Image Classification and more!

Today we announce the release of ML.NET 1.0 RC (Release Candidate) (version 1.0.0-preview ), which is the latest preview release before the release of the final version of ML.NET 1.0 RTM in Q2 2019.

Soon we will complete the first milestone of an interesting development that began in May 2018, when we released open-source ML.NET 0.1. Since then, we have released 12 preview releases (one per month), as shown in the roadmap below:

image

The first thing we did in this release (ML.NET 1.0 RC) was to finally complete the basic edits in the API. In the next sprint, we will focus on improving documentation and examples, as well as addressing key critical issues, if necessary.

The goal is to avoid any new critical changes as you move forward.

ML.NET 1.0 RC Timeframe Update


  • The separation of the Stable and Preview versions of ML.NET: ML.NET 1.0 and the main part of the ML.NET functionality (about 95%) will be released as a stable assembly of Stable (version 1.0).

    You can find a list of links to stable builds here .

    However, there are several feature sets that will still not be in RTM state with the release of ML.NET 1.0. These are features that still retain Preview status. They will be released in 0.12.0-preview .

    The following are the main features that will continue to work in preview mode after the release of ML.NET 1.0 ( 0.12 version packages ):

    1. TensorFlow Components
    2. Onnx Components
    3. TimeSeries Components
    4. Recommendation Components

    You can view the full list of links to Preview-assemblies “after 1.0” (0.12.0-preview) here .
  • IDataView moved to Microsoft.ML namespace: One of the changes in this release is that, based on our feedback, we moved IDataView back to Microsoft.ML namespace.
  • TensorFlow Support Enhancements: TensorFlow is an open source machine learning system used for deep learning scenarios (such as computer vision and natural language processing). ML.NET supports the use of TensorFlow models, but there were several issues with ML.NET version 0.11 that were fixed in RC version 1.0.
    You can view sample ML.NET code using the TensorFlow model here .
  • ML.NET 1.0 RC Release Notes: You can find additional release notes for 1.0 RC here .

Critical Changes in ML.NET 1.0 Release Candidate


For your convenience, if you move your code from ML.NET v0.11 to v0.12, you can see a list of critical changes .

Are you planning to go into production?


image
If you use ML.NET in your application and want to go into production, you can talk to the engineer from the ML.NET team to:

  • Get help in successfully implementing ML.NET in your application.
  • Leave feedback about ML.NET.
  • Demonstrate your application and possibly post it on your ML.NET homepage, .NET blog, or other Microsoft channel.

Fill out this form and leave your contact information at the end if you want someone from the ML.NET team to contact you.

Get Ready For ML.NET 1.0


image
As already mentioned, ML.NET 1.0 is almost ready! You can prepare for the release by

exploring the following resources: Start exploring ML.NET here .

Further, in depth, study some other resources:

You can leave feedback with any questions, suggestions or improvements in the ML.NET repository on GitHub . It will help us a lot to improve ML.NET and make .NET a great platform for machine learning.

Thanks and happy coding with ML.NET!

ML.NET team.

Also popular now: