Visual Studio AI Offline Model training (WINDOWS10)

First, the order

Environment construction: "View"
In the environment building process, through the Git clone gets to the Samples-for-ai folder, its subfolder examples directory is as follows.

Take TensorFlow as the sample model.
Reference blog: "Blog 1", "Blog 2".

Second, TensorFlow

(a) What is it?
Jane Book blog: view;
TensorFlow Chinese community: view.
is probably a Google open-source AI system.
(ii) vs Run offline model
Enter Samples-for-ai\examples\tensorflow, double-click to select TensorflowExamples.sln (Red a) to enter the project.

Select "Solution Explorer" (Red 1) in the right column of VS, find "MNIST" (Red 2), right click, select "Set as Startup Project" (Red 3),

Right-click the convolutional.py file in the Mnist project (Red 1) and select "Start Without Debugging" (Red 2). Such as

When my machine first runs, figure a appears after two minutes, and Figure B appears after five minutes.


It looks like the AI environment has been set up by a, B, but not yet. according to "blog 2", the Mnist folder should have more input, output, and export three folders after training, and these three folders will store the inputs of the training model, the checkpoint files at training and the final exported model files. I entered ai\samples-for-ai\examples\tensorflow\mnist again and did not find the above three folders. The contents of the Mnist folder before and after training C.

Find the answer ... Pending additions

Third, the matters needing attention

A Python environment
Some machines exist on multiple versions of Python and need to specify a python version.
Open Visual Studio, menu item "Tools" (Red 1)---> select "Python" (Red 2)---> select "Python Environment" (Red 3). Such as.

Appear. Red 1 indicates that the current Python version is 3.6, and Red 2 indicates that the Python version can be set to 2.7.

Visual Studio AI Offline Model training (WINDOWS10)