Machinelearning: Upgrade TensorFlow to 2.3 / Tensorflow.NET to 0.20
Hello,
I'm opening this issue for our discussion around migrating our implementation of Image Classification to TF.NET 0.20. It's release is imminent and wanted to have a public place for the discussion.
Cheers,
Jake
JakeRadMSFT
All 3 comments
JakeRadMSFT
on 18 Sep 2020
Hi @JakeRadMSFT, I'm so exciting to hear that ML .NET will upgrade its DNN module to align with lastet version of Tensorflow.NET. Current release is Preview 4 in NuGet. I'm working on fixing the last 3 examples, it will be pretty soon to release v0.20. We've already shipped tensorflow native 2.3 libraries include Windows, Linux and Mac as well as GPU version.
TF .NET v0.20 will bring a lot of new features, the some main things are EAGER MODE, tf.dataset data preprocess and Keras high level deep learning framework which is partially working.
Oceania2018
on 18 Sep 2020
@harishsk @JakeRadMSFT
Should we update tf document since the version of tf.net and tf runtime has changed.
LittleLittleCloud
on 13 Oct 2020
Related issues
lionelquirynen
路
3Comments
dev8546
路
3Comments
rogancarr
路
4Comments
frankhaugen
路
3Comments
sfilipi
路
4Comments
Most helpful comment
Hi @JakeRadMSFT, I'm so exciting to hear that ML .NET will upgrade its DNN module to align with lastet version of Tensorflow.NET. Current release is Preview 4 in NuGet. I'm working on fixing the last 3 examples, it will be pretty soon to release v0.20. We've already shipped tensorflow native 2.3 libraries include
Windows,LinuxandMacas well asGPUversion.TF .NET v0.20 will bring a lot of new features, the some main things are
EAGER MODE,tf.datasetdata preprocess andKerashigh level deep learning framework which is partially working.