sagemaker quality of life improvements

[oegedijk]

[oegedijk]

@absynthe started a new PR with the quality of life improvements. BTW, do you maybe know how to detect whether we're in a sagemaker environment?

[oegedijk]

so something similar to

ipython_kernel = str(get_ipython())
self.is_notebook = True
self.is_colab = True if 'google.colab' in ipython_kernel else False

[oegedijk]

@absynthe I've been toying around with sagemaker studio today, but so far haven't been able to get the dash to run, do you have a guide or HOWTO?

[absynthe]

Hey @oegedijk, I created this working example for you: https://github.com/absynthe/explainer-dashboard-studio-working-example/blob/main/workingExplainerDashboardStudioExample.ipynb . Try to walk through it and let me know if you have any questions.

[absynthe]

so something similar to

ipython_kernel = str(get_ipython())
self.is_notebook = True
self.is_colab = True if 'google.colab' in ipython_kernel else False

There is nothing in the kernel giving SageMaker studio away. However, you can look for the /opt/ml/metadata/resource-metadata.json file in the file system.

[absynthe]

Hi again @oegedijk , I added a test for kwargs: https://github.com/absynthe/explainer-dashboard-studio-working-example/commit/70668866f3bfb5e48bd495f316dfb3a1d4a4f5e2 . This does not seem to work, I assume something is missing for the cli glue to work?

We cannot run the dashboard directly in the notebookn using the run command. This is due to the separation between compute and storage of SageMaker Studio as mentioned here: https://aws.amazon.com/blogs/machine-learning/dive-deep-into-amazon-sagemaker-studio-notebook-architecture/ .

This is why there is the workaround of exporting the dashboard config to disk and loading again in the terminal window of the JupyterServer App.

[DannyDannyDanny]

We'd very much like to start using explainer dashboard on SageMaker. Is there anything I can do to help push this forward?