Machinelearningnotebooks: RunDetails widget doesn't appear

[tinkerer]

Using jupyter notebook 5.7.4 with Python 3.6.8 and running through the azureml examples. The RunDetails Widget does not render when calling show().

For example, this is what I see when I attempt to show the RunDetails in the train-hyperparameter-tune-deploy-with-tensorflow example:

[ahcan76]

This looks like the extensions are not enabled. could you provide output of:
jupyter nbextension list

[archanpi]

Is this still an issue?

[tinkerer]

No extensions come up when I run that command:
"C:\Users_____>jupyter nbextension list
Known nbextensions:

"
This is still an issue. What do I need to do to enable extensions? Are extensions enabled by default if you run through the installation using anaconda?

Otherwise, I don't see any information about enabling extensions prior to using the RunDetails widget in any of the documentation where such a clue might belong:

https://docs.microsoft.com/en-us/azure/machine-learning/service/quickstart-create-workspace-with-python
https://docs.microsoft.com/en-us/azure/machine-learning/service/tutorial-train-models-with-aml
https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-track-experiments

[swanderz]

@archanpi, I see the same behavior as @tinkerer and I below is my output.

Does it indicate that I don't have the azureml_train_widget in my azure_automl env?

(azure_automl) c:\Users\anders.swanson\Documents\attrition>jupyter nbextension list
Known nbextensions:
  config dir: C:\Users\anders.swanson\.jupyter\nbconfig
    notebook section
      azureml_train_widgets/extension enabled
      - Validating: ok
      azureml_widgets/extension enabled
      - Validating: ok
  config dir: C:\Users\anders.swanson\AppData\Local\Continuum\miniconda3\envs\azure_automl\etc\jupyter\nbconfig
    notebook section
      azureml_widgets/extension enabled
      - Validating: ok
      jupyter-js-widgets/extension enabled
      - Validating: ok
      nb_conda/main enabled
      - Validating: ok
    tree section
      nb_conda/tree enabled
      - Validating: ok

[archanpi]

@tinkerer It looks like you're not running this in conda.
Run this command in the conda environment
jupyter nbextension list

[archanpi]

@swanderz It looks like you have some deprecated user widgets installed. You should first uninstall them by running this-
jupyter nbextension uninstall --py --user azureml.widgets
jupyter nbextension uninstall --py --user azureml.train.widgets

If the second command throws an exception, you can ignore it.

See if this works for you. If this doesn't work, send the console error from the browser

[archanpi]

@tinkerer Also, #4259 issue doesn't seem to be connected to this issue.

[tinkerer]

I reinstalled following the conda instructions on the quickstart for python. First executing the RunDetails line complains about azureml.widgets not being installed. RunDetails widget still doesn't appear after pip install azureml.widgets. Heres what comes up when I run !jupyter nbextension list

[archanpi]

Your environment looks good. Try restarting the kernel and refresh the browser. Also, check if the Developer Console has this message displayed - "Loading extension: azureml_widgets/extension"

[swanderz]

@archanpi uninstalled the extension like you said and I can now see the the widget in Jupyter Notebook.

To confirm: this extension will not work in either:

• Jupyter Lab, or
• VSCode Python extension's Jupyter server

[archanpi]

Excellent!
Jupyter Lab - not working currently
VSCode - no, it never worked

[swatig007]

closing this issue since uninstalling old widget and reinstalling it fixed the issue

[scign]

Just to share my situation: I had jupyter running in one conda environment but using a kernel from a different environment. The azureml-sdk[notebooks] package has to be installed in the environment where jupyter is installed, not the kernel you're running.

[swanderz]

@scign good to know! conda is amazing, but quirks like this can be maddening...

[scign]

@scign good to know! conda is amazing, but quirks like this can be maddening...

More of a thing about jupyter environments than conda! It allows you to run kernels in other environments which is awesome but the widgets need to be installed in the environment in which you're running the notebook, not the environment in which you're running the code - which makes sense really. Definitely good practice to separate the two so it's a "quirk" I'm willing to live with.