Created a Computer Vision project with the VS Tools for AI extension after creating the service in the Azure Portal. Putting the API key and (I assume) the correct location, I'm still getting an unauthorized exception. I even copied both keys from the AI Tools section of the Server window and still get this error. Am I setting something wrong?

I tried multiple attempts and kept having this error, asking me to have account name to be between 3 to 32 characters long. But, the aAccount name is already between 3 to 32 characters--DrDataSciExperiment. preciate if you can help. Thank you.

Messages below is from output window.

Retrieving resource groups ...
AMLCli: Executing 'az group list -o json'.
AMLCli: Execution succeeded.

Done.
Retrieving accounts ...
AMLCli: Executing 'az ml account experimentation list -g "AI1" -o json'.
AMLCli: Execution succeeded.
Done.
Retrieving workspaces ...
AMLCli: Executing 'az ml workspace list -a "DrDataSciExperiment" -g "AI1" -o json'.
AMLCli: Execution succeeded.
Done.
AMLCli: Executing 'az ml project create -n ClassifyingIris_074 -u "https://github.com/Azure/MachineLearningSamples-Iris" --path C:\MachineLearn -a "DrDataSciExperiment" -w "DrDataSciWorkSpace" -g "AI1" -l "australiaeast" -o json'.
ERROR: Account name must be between 3 and 32 characters long. Its first character has to be alphanumeric, and the rest may contain hyphens and underscores.

This repository is currently missing a LICENSE file.

A license helps users understand how to use your project in a compliant manner. You can find the standard MIT license Microsoft uses at: https://github.com/microsoft/repo-templates/blob/main/shared/LICENSE.

If you would like to learn more about open source licenses, please visit the document at https://aka.ms/license (Microsoft-internal guidance).

When a Tensorflow model .pb file is imported using the Visual Studio Tools for AI import wizard, if the "Optimize" checkbox is checked, when executed, the call to manager.RunModel will run, but never finish.

My example is a very simple network built with Keras...
Conv2D(64, kernel_size=3, activation='relu', input_shape=input_shape),
Dropout(0.2),
Conv2D(32, kernel_size=3, activation='relu'),
Dropout(0.2),
Flatten(),
Dense(num_classes, activation='softmax')])

Running ONNX 1.3, Tensorflow 1.10.0.

Just trying to create a new AI Tools Inference project from template, File > Project > AI Tools > Inference.

Fails with message:
Module tensorflow missing
Please visit https://go.microsoft.com/fwlink/?linkid=872929 to prepare the development environment.

Tensorflow seems to be installed, see pic.

Please advise, thanks in advance!

When I submitting a job, it comes out the error message: The terminal process command '/bin/zsh ~/.vscode/extensions/ms-toolsai.vscode-ai-0.1.9/out/scripts/pythonTerminal.sh' failed to launch (exit code: 127).

And when I type the above command in the command line, it showed me "pythonTerminal.sh" did not handle directory paths with spaces correctly.

Like, "/Applications/VMware Fusion.app/Contents/Public" seems to miss a backslash.

"Input directories" and "Output directories" list lists of Submit job dialog show entries as $AZ_BATCHAI_MOUNT_ROOT_. The should be $AZ_BATCHAI_INPUT_ and $AZ_BATCHAI_OUTPUT_ correspondingly.

Product current does not detect CUDA 10 with latest cnDNN

[WARNING] [Microsoft Visual Studio Tools for AI] Not detect CUDA! We recommend CUDA 9.0 (https://developer.nvidia.com/cuda-toolkit). The installer script will install dependency package for CUDA 9.0 by default.

It shows "Error - ssh failed, exit code: 123" after I successfully submited my job. However, I can ssh my VM on my terminal. And it shows "exit code: 127" in the same situation sometimes.

Attempting to follow the installation instructions to install this package from within VS2017 fails - the package doesn't appear to be available when searching for it in the Extensions and Updates dialog:

The VSIX package is available for download, however, from the Visual Studio Marketplace.

I have used AI Tools to generate a inference model with my python model, i want to use C# call for it.
However, when i run it, it occur to the following problem. My build platform has been changed to x86, but it doesn't work.
I don't know how to solve it, can you help me?

Great to see Microsofts commitment to AI. MS has always pioneered the idea that serving developers should be a huge priority.

And as much as I appreciate that you're supporting Python and Tensorflow as they are the industry standard, I'd love to see much more integrated tools with CNTK .NET in a way to draw in the average c# developer. There is a huge potential here.

Another thought...
Just as 3D rendering used to be a developers task (remember Pov Ray 😄) and is now all achieved through very comprehensive editors.
I'm confident AI will go the same way, so get your ducks in a row ;)

Thanks for the great work you're doing here!

Tried installing today after updating VS to latest 15.8.2, but the extension says it is missing a prerequisite. In digging into the logs I found it's failing due to Microsoft.VisualStudio.Component.Common.Azure.Tools:

8/29/2018 4:05:48 AM - The extension with ID 'ec62d75b-40eb-4fad-806c-c8196b2bd091' is not installed to Microsoft Visual Studio Enterprise 2017.
8/29/2018 4:05:50 AM - Extension cannot be installed to the following products due to missing prerequisites:
8/29/2018 4:05:50 AM - 	Microsoft Visual Studio Enterprise 2017
8/29/2018 4:05:50 AM - 		-------------------------------------------------------
8/29/2018 4:05:50 AM - 		Identifier   : Microsoft.VisualStudio.Component.Common.Azure.Tools
8/29/2018 4:05:50 AM - 		Name         : Connectivity and publishing tools
8/29/2018 4:05:50 AM - 		Version      : [1.10,2.0)
8/29/2018 4:05:50 AM - 		Error        : The prerequisite version specified does not match the version installed
8/29/2018 4:05:50 AM - 
8/29/2018 4:05:50 AM - System.InvalidOperationException: No modifications could be applied.

I'm not entirely sure how to check the version I have installed, but both the VS installer and the VS extension manager say I have everything fully up to date, including everything Azure-related.

Looks like the description strings got corrupted

I'm trying to import an ONNX model generated by the Cognitive Service - Custom Vision - Classifier . I've got the following output log in the AI Tools window:

C:\Users\frank\AppData\Local\amlworkbench\Python\lib\site-packages\h5py\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters
Traceback (most recent call last):
  File "C:\USERS\FRANK\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_15C46C60\EXTENSIONS\JIZDA1Z2.1ND\RuntimeSDK\mlscoring\exporter\model_preprocess.py", line 503, in <module>
    main()
  File "C:\USERS\FRANK\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_15C46C60\EXTENSIONS\JIZDA1Z2.1ND\RuntimeSDK\mlscoring\exporter\model_preprocess.py", line 59, in main
    pre_process(preprocessParams, args.result_path)
  File "C:\USERS\FRANK\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_15C46C60\EXTENSIONS\JIZDA1Z2.1ND\RuntimeSDK\mlscoring\exporter\model_preprocess.py", line 80, in pre_process
    pre_process_onnx_model(preprocessParams, result_path)
  File "C:\USERS\FRANK\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_15C46C60\EXTENSIONS\JIZDA1Z2.1ND\RuntimeSDK\mlscoring\exporter\model_preprocess.py", line 113, in pre_process_onnx_model
    interfaces,_,_,_= extract_onnx_model_information(preprocessParams.src_path)
  File "C:\USERS\FRANK\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_15C46C60\EXTENSIONS\JIZDA1Z2.1ND\RuntimeSDK\mlscoring\exporter\onnx_exporter\onnx_exporter.py", line 197, in extract_onnx_model_information
    sn_type = _onnx_type_to_mlscoring_type(data_type_name)
  File "C:\USERS\FRANK\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_15C46C60\EXTENSIONS\JIZDA1Z2.1ND\RuntimeSDK\mlscoring\exporter\onnx_exporter\onnx_exporter.py", line 211, in _onnx_type_to_mlscoring_type
    ERROR_CODE["UNSUPPORTED_DATA_TYPE_ERROR"][1].format(onnx_type))
mlscoring.exporter.exception.ExportException: 208:'Using unsupported tensor data type UNDEFINED'

When I am run mnist.py then
TypeError("int() argument must be a string, a bytes-like object or a number, not 'NoneType'",)

HI there

are there any c# examples of how to correctly pass a image across to a tensorflow model in ai tools after the Model project has been made?

cheers

Hi, when can we expect the VS2019 version, please?

Thanks

please consider adding support for frozen models when using the model inference tool (given that this support is already present in the model converter)

I downloaded the tools and installed it, however it was a pain to startup visual studio as it was causing it to freeze. After uninstalling the problem seems to have gone away.

In Server Explorer, refreshed, the AI tools are shown. Cognitive Services in my subscription for instance are listed. When right-clicking Azure Batch AI and then selecting "Add Cluster" it will open a dialog.

This Dialog ("Add Cluster") is not loading or showing the resources groups in my subscription, thus I cannot create a cluster.

I am using the latest AI Tools and Vis Studio 2017 as of today.
AI Tools Version: 0.4.0.620119

best regards, Robert

Webpage is unreachable when running TensorBoard (http://127.0.0.1:1025)

The issue was originally posted here: https://developercommunity.visualstudio.com/content/problem/266122/vs-158-preview-2-uses-guid-as-class-name-for-onnx.html

Even after changing the onnx file name, VS 15.8 preview 2 uses a guid as a class name for ONNX models exported and downloaded from customvision.ai

Steps to repeat:

• Set customvision.ai project type to general (compact) domain
• Upload images and train model
• Export and download ONNX model
• Import ONNX model in Visual Studio

Expected result: allow user to enter a readable name or use the project name from customvision.ai instead of a guid as a classname

Issue:
When submitting a Python job, Tools for AI showed error message that some {GUID}.tar.gz package couldn't be found in %TEMP%\8\ToolsForAITemp.
This folder path was also set in VS Options as temporary directory of AI Tools.

After resetting AI Tools option, the 8 in temporary directory above became 4, and it was possible to submit Python job.

Environment:
OS: Windows Server 2012 R2 Standard
IDE: VS2017 15.8.0

Describe the bug
I use Microsoft Visual Studio 2017 Community Edition with Turkish Language Pack and downloaded Microsoft AI Tools. The Tools for AI tab in Visual Studio application is in Chinese...

Expected behavior
The tab must be in Turkish.

Screenshots
https://imagizer.imageshack.com/img921/4471/PZnow6.png

Desktop (please complete the following information):

• OS: Microsoft Windows 10
• Version Microsoft Visual Studio 2017 Community Edition + Microsoft Tools for AI

I have explicitly set my Environment.Startup settings to "Show empty environment".
Yet the "Tools for AI Start Page" always pops up when I start the VS 2017 IDE.

Most of my workload is not related to the AI tools.
Why does this package even load into the IDE at startup time?

Extensibility packages are supposed to load on demand, and not at startup. At a bare minimum, the "Show empty environment" at startup, should be honored. But in reality, you probably shouldn't even have your package loaded during startup.

Running the Visual Studio Tools for AI Inference project import wizard, importing a .pb file, the wizard gives a warning saying version 1.3 of ONNX is not supported, and version 1.10.0 of Tensorflow is not supported, even though they are supported according to this page on Microsoft.ML.Scoring 1.2.

Do you have any plan for Mac support?

the model.dll have a exception.

Hi everybody

Visual Studio 2017 15.7.5

AI Tools 0.4.1.664470

Windows 10.

Every time i try and create a new AI Tools -> "Module Inference Library (.Net Framework)" project i get a "Python module is missing!"



error and i cant make the project. Which isnt great.

Python is installed and i can create a python project /application.

How do we fix this?

cheers

issue reproduce and question redirected from here

I have a very simple Cognitive Services app that's trying to generate a thumbnail.

var imageUrl = "https://shop.mrmen.com/wp-content/themes/mrmen/assets/characterpic-square-mr%20small.jpg";
var vision = new ComputerVisionAPI(new ApiKeyServiceClientCredentials(key));
vision.AzureRegion = AzureRegions.Westcentralus;
var response = await vision.GenerateThumbnailWithHttpMessagesAsync(600, 400, imageUrl, smartCropping: true);

But even though the URL is plainly accessible to me, I get the following error message.

{"code":"InvalidImageUrl","requestId":"d5c8e9ef-f85f-4ee9-92d1-23fdda9cf2cc","message":"Image URL is not accessible."}

The only obvious difference I see is the '%'. Maybe it isn't being encoded correctly? Can you investigate?

I have installed newest AI tools version 0.3.3.0 on VS 2017 (15.5.6). When I open an existing project I get the error "Save view failed with the following error: Type 'Microsoft.VSDesigner.ServerExplorer.Node' in assembly 'Microsoft.VisualStudio.ServerExplorer, Version=15.0.0.0, Culture=neutral, PublicKeyToken=b03f5f711d50a3a' is not marked as serializable. Make sure the view name is a valid file name.

If I disable AI tools the error does not appear again.

Hi everyone,

I'm facing some errors when I try to convert or import a tensorflow .pb model using AI Tools converter or Import Model. The message that I've gotta is this:

KeyError: "The name 'input:0' refers to a Tensor which does not exist. The operation, 'input', does not exist in the graph."

I'm passing the correct internal names for input and output tensors (I'm able to load and use the same model inside a Python script). I've visualized my model using Netron and 'input' exists in the graph.
The model is available here.

Thanks in advance.

There are important files that Microsoft projects should all have that are not present in this repository. A pull request has been opened to add the missing file(s). When the pr is merged this issue will be closed automatically.

Microsoft teams can learn more about this effort and share feedback within the open source guidance available internally.

Merge this pull request

HI, I want to convert a scikit-learn model to onnx,but I dont now how to set up the input_featrues parameters. thanks :)

my model featrues :
Sepal Length float 1
Sepal Width float 1
Petal Length float 1
Petal Width float 1

Here are the error logs :

System.Exception: D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\sklearn\base.py:311: UserWarning: Trying to unpickle estimator LogisticRegression from version 0.18.1 when using version 0.19.1. This might lead to breaking code or invalid results. Use at your own risk.
UserWarning)
Traceback (most recent call last):
File "C:\USERS\H00316469\APPDATA\LOCAL\MICROSOFT\VISUALSTUDIO\15.0_04A81442\EXTENSIONS\Y2LMQEZR.VBA\RuntimeSDK\model\model_converter_cli.py", line 203, in convert_sklearn_to_onnx
model_onnx = convert_sklearn(model_sklearn, extra_args.get(paramname_graph_name), input_features)
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\winmltools\convert\main.py", line 20, in convert_sklearn
return _convert_sklearn(model, name, input_features)
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\onnxmltools\convert\sklearn\convert.py", line 109, in convert
nodes += _convert_sklearn_node(context, model, node_inputs)
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\onnxmltools\convert\sklearn\convert.py", line 159, in _convert_sklearn_node
return _do_convert(context, converter, node, inputs)
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\onnxmltools\convert\sklearn\convert.py", line 148, in _do_convert
node = converter.convert(context, sk_node, inputs)
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\onnxmltools\convert\sklearn\GLMClassifierConverter.py", line 63, in convert
class_labels = utils.cast_list(int, classes)
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\onnxmltools\convert\common\utils.py", line 144, in cast_list
return [type(item) for item in items]
File "D:\sofatware\microsoft Visual Studio\shared\Python36_64\lib\site-packages\onnxmltools\convert\common\utils.py", line 144, in
return [type(item) for item in items]
ValueError: invalid literal for int() with base 10: 'Iris-setosa'

Best regards

很有缘能看到你的文章：《AI应用开发实战 - 从零开始配置环境 》。感觉思路清晰，是在行业内深耕多年的专业人士。目前我们有一个项目在开发，关于chat GPT 内核的应用。如果愿意合作，请给我打电话：13615899736，或回复邮件。

VSIXInstaller_c7388073-99e5-4c83-b305-9b3b49e167ad.log

Issue:
After submitting a Python job to remote Linux machine, VS crashes when opening Jobs view in Server Explorer - AI Tools - Remote Machines - {SomeUbuntu} - Jobs.
After restarting VS, job submitted before crashing is not listed in Jobs view.

Environment:
OS: Windows Server 2012 R2 Standard
IDE: VS2017 15.8.0

I've completed the last step by installing the Tools for AI plug-in, but I haven't found the AI-Tools option in the Visual Studio menu bar.Please do not hesitate to give your advice.

Hi. I'm trying VS AI tool to convert model but dumps v4 op vs. v7. Is there update to AI tool?

microsoft/Windows-Machine-Learning#50
No suitable kernel definition found for op Add issue

Thanks

Hi all,
I'm facing a few issues with the Inference project wizard.
The first one does not seem to be critical : using VS17, ML.Scoring 1.2.0, Tensorflow 1.9.0 or 1.10.0 and Onnx and 1.2.3 or 1.3.0 : on getting the Python Environnement, the wizard triggers the two following errors :

• Has detected that Onnxversion 1.2.3 is higher than that Microsoft.ML.Scoring library accepts. (either 1.2.3 or 1.3.0)
• Has detected that Tensorflow version 1.10.0 is higher than that Microsoft.ML.Scoring library accepts. (no error if I use TF 1.9)
  However this does not prevent me from building the Inference library.

Now back to my second and main issue : evaluation values do not match evaluation results in Python with ONNX model converted from Keras.
Edit : problem solved... beware of image size tuple convention in Python : (height, width) and not (width, height) as I would have expected, as in C# for imaging components. So I trained the CNN with the wrong image input dimensions, and fed the C# program with something the network didn't expect. My bad.

FYI, I tested a model compiled and trained with TF 1.12.0, it seems to work !

I imported the onnx model in the project and when I run it ,a issue has occur.
System.Runtime.InteropServices.SEHException
HResult=0x80004005
Message=外部组件发生异常。
Source=<无法计算异常源>
StackTrace:
<无法计算异常堆栈跟踪>
what's wrong?