From @fucknamel on Auguest 14, 2018

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.

Copied from original issue: microsoft/vs-tools-for-ai#27

I am unable to open a jupyter notebook with an environment variable in my python path: ${env:HOME}/.../python
The extension cannot find jupyter.
The extension works correctly when I change the path to a normal one.

Hi, I use VS code to submit a test script to one of my remote linux machine. After submitting the job, I click the button "List Jobs", it shows that "Error - Can not parse list job result".

The terminal shows:
Parse job list faild:

Error: Unexpected end of JSON input

But when I change to submit jobs to another linux machine, there is no error shows up.
How can I debug that?

OS: OS X
VS Code: 1.20.1
Extension: 0.1.7, Feb 9 2018, public preview

STEPS TO REPRODUCE:

1. Open VS Code
2. Open Azure ML Sample Explorer
3. Select install button for 'Simple Linear Regression'
4. Hit enter when prompted to input the project name (default is simplelinearregression)
5. Select account prompt appears (sometimes). When it does, it only lists two accounts, neither of which are mine (I'm on a shared, internal MSFT account)

'Account' may be the wrong wording to use in step 5. It's actually looking for a Machine Learning Experimentation. If you go to the Azure Portal, you can see the name of the item.

IDEAL SCENARIO:

• Ability to create a new experiment from the wizard
• Rename "account" in step 5 to Machine Learning Experimentation

It will often hang when trying to retrieve a list of accounts. On occasion it will work, but will only return two works paces when it does.

I've verified that I have logged into Azure via the command line, and selected the appropriate account, subscription id# 3fee811e-11bf-4b5c-9c62-a2f28b517724 This is the same account that I use in ML Workbench, and it works fine.

Installed vstools for ai and now when i try to open a sample azure ml , i get this error: “command 'vscodeai.openGallery' not found”

Hi all! When trying to install in my x64 instance of VS Code, I get an unsupported version error. I presume - though I am not entirely sure - that it is due to the 64 bit version. If that's the case, a 64-bit build would be appreciated :)

I follow the instruction step and it also give me the sign that --> verbose: Jupyter server started in the output panel , but I can't see any ipynb file content, somebody got the same problem ?

I install Azure Machine Learning Workbench and go to find "ML Sample", but when I install I receive as follow:

=== Create Project From Gallery ===
Please input the project name
Please input the project installation folder
Getting account list....
Create project from gallery failed, error=Error: Can not get existing accounts, please create one account and try later.

I already login, anyone could give me some suggestion?

Thank you !

Latest VSCode TfAI installed / running locally on RHEL Server release 7.5 (Maipo)
When I start VSCode I get this message box in the lower-right-hand corner of the screen:
(represented below as ascii art where (x) and *, v, and x represent buttons/icons).

+-----------------------------------------------------------------------------------------+
| (x)  Initialize log file failed: Path must be a string. Received undefined.     *  v  x |   
|                                                                                         |
|Source: Visual Studio Code Tools for AI (Extension)                                      |
+-----------------------------------------------------------------------------------------+

This was a headache to debut but I narrowed it down to this extension. I suddenly noticed all of my syntax highlighting for comments with '--' broke. It would colour as regular text so any keywords inside and any single quotes (also apostrophes) would not end and break the entire document. I removed extensions one by one until I found this was the root cause.

For anyone searching on Google, SQL, highlight, comment, escape, single

When open ipynb file "View in Jupyter Server" got this error "Logger failed : Cannot read property 'log' of undefined"
OS: Linux Mint 19
python: system-3.6.5

The Tools for AI extension is often installed as part of the Azure Extension Pack.

Now imagine if every extension part of this pack had a Start Page that opened automatically. It would give a pretty bad experience to users.

Currently however only the Tools for AI extension does this. But it does this on every startup of VSCode by default. This behavior can be disabled by unchecking an option on the start page, but this option is not visible at first glance. You have to scroll all the way down the start page to find it.

Would it perhaps be possible to implement one of the following changes?:

• Disable opening the start page on startup by default when the extension is installed via the Azure Extension Pack.
• Make the option to disable this behavior visible at first glance from the start page (without scrolling).

I am not able to locate the "ML Sample" command after installing these tools.
I am on windows 10 64 bit version and running visual studio code.

I have Jupyter installed on my computer (macOs Sierra 10.12.6) and I would like to specify the location path to the extention, in order to be able to run the notebooks.

When I right-click AI: view in Jupyter Server, I get the error error: Jupyter is not installed in the current python path: /usr/local/bin/python3, please install jupyter first., which clearly shows that VSCode is not looking into the right path to start it: how can I custom-specify the Jupyter path?

When loading up jupyter-notebook.exe with python installed in a virtual environment on windows it cannot find the executable because it is looking for <PythonVenvDirecory>\Scripts\jupyter-notebook.exe

for the virtual environment the python executable is in the same directory as the jupyter-notebook.

as a workaround, from my python scripts directory I ran:
mklink /D Scripts .

A symbolic link causes the plugin to see the jupyter-notebook executable on the aforementioned path, and it works.

Hi, I want to open ipynb file in Jupyter server but show nothing on vscode.
Is there anything that I should do? My host OS is Win10.

This extension uses version 0.0.17 of vscode-extension-telemtetry module.

The latest version of vscode-extension-telemtetry module (0.1.0) uses the latest version of applicationinsights module (1.0.6) which has the fix to prevent unnecessary patching of the require and other 3rd party modules.

For more details, please see microsoft/vscode-extension-telemetry#16

I am using VS Code to submit jobs to a vm but there only seems to be one configuration file ai_job_properties.json. Is there a way to define multiple job_properties and select which one I wish to run?

When opening a .ipynb file in VS Code, it should by default open in the Jupyter server view, rather than opening it as a raw text file. Since 100% of users that want to open/interact with a .ipynb file, want to do so in the Notebook environment.

I am using VS Code to submit jobs to a vm, but the log/stdout only has preview, there is no idea to watch the realtime log. I think it is a good function.

After clicking AI: View in Jupyter Server, the system is not able to open the server, throwing the following console log:

Jupyter is not installed in the current python path: venv/bin/python, please install jupyter first.

I have correctly installed Jupyter, and it is possible to find the folder in the directory: venv/bin/python/jupyter-notebook

Hi, I'm using VScode on macOS.

When I submit the job(AI: Submit 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("/bin/zsh ~/.vscode/extensions/ms-toolsai.vscode-ai-0.1.9/out/scripts/pythonTerminal.sh") in the terminal, 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
---- "/VMware\ Fusion.app/".

with vscode, show following error:
Initialize log file failed: Path must be a string. Received undefined

Ubuntu 18.04 LTS
VSCode: 1.26.1

I created an experiment subscription and installed the Azure workbench, but creating a project from the sample explorer fails. Here is the output:

=== Create Project From Gallery ===
Please input the project name
Getting account list....
Getting workspace list....
Please select the project installation folder
Creating project....
Create project from gallery failed, Execute az ml project create -n "mnistusingtensorflow" -u https://github.com/Azure/MachineLearningSamples-tf -a "purnam-experiment" -w "PurnamWorkspace" -g "purnaj" -l "australiaeast" -o json failed, ERROR: Account name must be between 3 and 24 characters in length and use numbers and lower-case letters only.

Version 0.1.4 of the extension is installed.

I notice that this project is open source and released under MIT license, so where can I get the source code of this extension? Thanks.

I use VS code for different projects and I noticed that this extension uses port 8080.
This unfortunately clashes with nodejs.
I could use another port but in ideal world I would like to configure this extension to use a different port.

Is that at all possible?

Whenever I click on List Jobs I see nothing. If I go to visual studio and point to the same vm I can see all the jobs. Is this a bug or am I missing something

Can't find ML Sample

Are there plans to add linux support to this extension?

I first used this nice extension in Window, it work well. So I installed it in my Ubuntu environment. Unfortunately, there was a problem after the installation:
Initialize log file failed: Path must be a string. Received undefined
And if I ignore the prompt to use, such as open a jupyter notebook by this extension, it will prompt such an error:
Logger failed : Cannot read property 'log' of undefined
Hope to get answers, thanks!

After submitting a job, the state of my job is completedfail.
The log of stderr is:
Traceback (most recent call last):
File "mnist/main.py", line 3, in
import torch
ImportError: No module named torch

It seems that the environment in the remote Linux machine is not configured.

When I click click here to run the script and Run as Administrator is selected VS Code executes command:

undefined "/usr/bin/python3" install.py

Which obviously throws an error undefined: command not found

The extensions executes:

// pythonPath is path to python configured in settings
// in extension code this variable is called t
this.terminal.sendText(`${j[l.platform()]} "${pythonPath}" install.py`)

to get the command to run and j[l.platform()] returns undefined.
j looks like this:

j = {
    win32: ".\\runAsAdmin.cmd",
    darwin: "sudo"
};

so if l.platform() returns something other than win32 or darwin - eg. linux the command is prefixed with undefined instead of sudo

From @leozhangjie66 on June 7, 2018 10:15

• VSCode Version:1.24.0
• OS Version:win7/win10

I get a error what is "canot read property 'filtyer' of undefined" when i use the Tools For AI in VScode 1.24.0.

Copied from original issue: microsoft/vscode#51336

I typically leave my az configure settings set to table, but I wasn't able to add configuration in VS Code.

When I switched to jsonc it worked.

When running "AI: Open Azure ML Sample Explorer" from the command palette, I get the message "Please install the azure ml workbench to enable azure ml feature of this extension.". Azure ML Workbench is indeed installed.

Conda environment names should be shown as part of the name of the environment. The only way for a user to distinguish between their different environments in VS code is to look at the absolute file path.

For example, in the screenshot below, i had the conda environment with the name "ml" but you wouldnt know which conda selection to pick for "ml" unless you looked through each of their absolute file paths.

This extension works with Azure, so it makes sense to add it to the Azure category, so that it shows up properly in the marketplace

In package.json, add:

  "categories": [
    "Azure"
  ],

Trying to execute "AI: Azure ML - Login", getting this message

The terminal process command `/usr/local/bin/fish --init-file /Users/vlza/.vscode-insiders/extensions/ms-toolsai.vscode-ai-0.1.1/out/scripts/azMlTerminal.sh` failed to launch (exit code: 1)

So, it's trying to execute "shell" script via the fish shell. Should be executed with "sh -c" or something.

Let me know if more information is needed.

Hi, I would suggest that we should be able to refresh the list of compute targets without restarting code.exe. For instance, right clicking on "AzureML - Run Configuration" could show an option "Refresh" so information is re-read from the project folder (in the case that you rename the compute target for example, or other changes that aren't necessarily launched from this menu but by another application or by the user browsing to the compute and runconfig files. Same for Batch AI, if I create a cluster from the portal, it would be nice to see the possibility to refresh the menu.

Hi,

It would be great for you to include a link to your GH Issues in the package.json. This will allow users to see a link for this in the marketplace and in VS Code.

"bugs": {
    "url": "https://github.com/Microsoft/vscode-wordcount/issues",
    "email": "[email protected]"
},

Additionally today you have no short description and that would improve display in marketplace search results.

"description": "Markdown Word Count Example - reports out the number of words in a Markdown file.",

Sean
VS Code team

When I've installed this extension, I was shown some page that had a link to "tun a script to set up your environment and install tools".
Unfortunately, I've restarted VS code after seeing that page. Now I cannot get to that page again.
It's not the "AI: Open Tools for AI start Page" - it's different. It also does not have the link to the environment setup script.
I've tried reinstalling the extension, but that page was never shown again.

Expected behavior: There is a way to view that first startup page again.

P.S. Maybe you should just merge it to the AI Tools Start Page.

Looks like there's something in this extension that causes the Microsoft Python Extension to crash.

See here for confirmation and further info:
microsoft/vscode-python#432 (comment)

Would like to resolve this, unfortunately the source is not available. Please let us know how we can help.

Hi,
Does this extension available in VS2107 windows edition?
If yes, how can I install it?

安装时出现的问题：
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.

已经安装，但是不知道为什么没有检测到

Hi,

It's the first time where I use this extensions and I don't understand how I can configure the extension correctly. I would like to run a PyTorch code on a remote machine, in local, it work perfectly.

Environment data

VS Code version: 1.19
Python Extension version: 0.91
Visual Studio Code Tools for AI version: 0.1.6
Python Version: 3.6.5
OS and version: OSX High Sierra

Remote server:

Python Version: 3.5.2
OS and version: Ubuntu 16.04 LTS

Actual behavior

When I run a job, I get the following error :

Expected behavior

Automatic connection to the server, I think.

Steps to reproduce:

• Add Platform configuration
• Set host, userName variables for a SSH connexion
• Right click of your python code and click on "AI: Submit Job"

Do you have examples or a documentation to configure the ai_job_properties.json file and a remote server ? Because maybe I have misconfigured the extension

Thank you

Dear team,

I am trying to run to browse a gallery (ctrl-shift-p, AI Sample Gallery). After I choose it, it says it is unable to load and then this command disappears from the menu

eg:
when typing tf dot, the auto complete window popups after 4-5 sec

my os is window 10 and vscode version is 1.26.0
the extension version is 0.1.9

Hi,
i do not see the possibility to submit a azure batch ai Job in VS Code. I installed the tools for ai extension. Also there is no Batch AI in the AI Explorer. I am running VS Code on Mac OS. What needs to be installed to see Batch AI in VS Code?

Thanks in Advance...
Kind regards,

Dirk

Create project using VSCODE " az --debug ml project create...." fails at:
requests.packages.urllib3.connectionpool : "DELETE /subscriptions//resourceGroups/mlbpgroup/providers/Microsoft.MachineLearningExperimentation/accounts//workspaces//projects/classifyingiris?api-version=2017-05-01-preview HTTP/1.1" 200 None

with response:
Response content:
msrest.http_logger : b'{\n "version": {\n "major": 1,\n "minor": 1,\n "build": -1,\n "revision": -1,\n "majorRevision": -1,\n "minorRevision": -1\n },\n "content": null,\n "statusCode": "OK",\n "reasonPhrase": "OK",\n "headers": [],\n "requestMessage": null,\n "isSuccessStatusCode": true\n}'

but ErrorMessage: Expecting value: line 1 column 1 (char 0)

Trying to create project in AML workbench shows error: Cannot clone remote Git repository.

Environment: Win10, fully patched, vscode and vscode-tools-for-ai fresh install

So neither approach will create a project!

From microsoft/vscode#44016 (comment)

When The Visual Studio Code Tools for AI is installed, SQL highlighting in VSCode changes.

The problem is that there's a clash of scope name:
"scopeName": "source.sql"

I agree that this should be handled better by VSCode, but it would good if you could change the scopes to something like source.cntk.