Adding the official version for SDK-FB-prophet-online-forecasting-notebook

[krishr2d2]

Checklist for moving the notebook to the main repo:

• Use the notebook template as a starting point.
• Follow the style and grammar rules outlined in the above notebook template.
• Verify the notebook runs successfully in Colab since the automated tests cannot guarantee this even when it passes.
• Passes all the required automated checks. You can locally test for formatting and linting with these instructions.
• You have consulted with a tech writer to see if tech writer review is necessary. If so, the notebook has been reviewed by a tech writer, and they have approved it.
• This notebook has been added to the CODEOWNERS file under # Official Notebooks section, pointing to the author or the author's team.
• The Jupyter notebook cleans up any artifacts it has created (datasets, ML models, endpoints, etc) so as not to eat up unnecessary resources.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[kweinmeister]

/gcbrun

[kweinmeister]

/gcbrun

[kweinmeister]

/gcbrun

[kweinmeister]

I saw a failure installing pystan in the build logs.

[krishr2d2]

I saw a failure installing pystan in the build logs.

It must be the build related error messages while installing the fbprophet library. Those messages while installation seem common while installing fbprophet according to : facebook/prophet#1598 but ultimately the library gets installed.

[ivanmkc]

@sudarshan-SpringML Please fix PR title. It's not clear if it's a small edit or a whole new notebook from the title.

[ivanmkc]

@krishr2d2 The installation seemed to never finish after 17 hours. This is one of those cases where you'll have to dig a bit into why pip is behaving this way.

I recommend using Vertex Workbench with a custom container to debug:

The image you should use is gcr.io/cloud-devrel-public-resources/python-samples-testing-docker:latest

[krishr2d2]

@krishr2d2 The installation seemed to never finish after 17 hours. This is one of those cases where you'll have to dig a bit into why pip is behaving this way.

I recommend using Vertex Workbench with a custom container to debug:

The image you should use is gcr.io/cloud-devrel-public-resources/python-samples-testing-docker:latest

@ivanmkc Are there any additional steps to get the instance started?? I tried creating the instance using the provided image but it doesn't allow me to access the JupyterLab on it.

The "OPEN JUPYTERLAB" button gets greyed out.

Alternatively, I've tested separating out the steps that install the fb-propphet's dependencies and the fb-prophet library without any installation errors on the user-managed instances but unable to run the linter test successfully on the changes. The linter test gave the following two errors :

error-1 : ImportError: cannot import name 'ParameterSource' from 'click.core' 

error-2 : ModuleNotFoundError: No module named 'absl'

These seem like coming from the linter test script rather than the the notebook.

[krishr2d2]

@krishr2d2 The installation seemed to never finish after 17 hours. This is one of those cases where you'll have to dig a bit into why pip is behaving this way.
I recommend using Vertex Workbench with a custom container to debug:
The image you should use is gcr.io/cloud-devrel-public-resources/python-samples-testing-docker:latest

@ivanmkc Are there any additional steps to get the instance started?? I tried creating the instance using the provided image but it doesn't allow me to access the JupyterLab on it. The "OPEN JUPYTERLAB" button gets greyed out.

Alternatively, I've tested separating out the steps that install the fb-propphet's dependencies and the fb-prophet library without any installation errors on the user-managed instances but unable to run the linter test successfully on the changes. The linter test gave the following two errors :

error-1 : ImportError: cannot import name 'ParameterSource' from 'click.core' 

error-2 : ModuleNotFoundError: No module named 'absl'

These seem like coming from the linter test script rather than the the notebook.

The linter test related errors are gone after re-installing the requirements for linter test.

[andrewferlitsch]

@ivanmkc gcbrun was 16 days ago, but still running?

[krishr2d2]

@ivanmkc , Is there a way to access the executed notebook for this PR. I can't find it currently as the logs are truncated. Also, the docker builds are replaced by gcloud build commands. Still, I suspect that the container building process is taking a lot of time in this case ultimately leading to a TIMEOUT.

[andrewferlitsch]

ValueError Traceback (most recent call last)
Step #5: Step #4: Cell In [21], line 6
Step #5: Step #4: 1 get_ipython().run_line_magic('cd', 'app/')
Step #5: Step #4: 3 # Choose a few product categories to build forecasts for from the Iowa Liquor sales data in BQ: bigquery-public-data:iowa_liquor_sales.sales
Step #5: Step #4: 4 # 1012100 CANADIAN WHISKIES
Step #5: Step #4: 5 # 1031100 AMERICAN VODKAS
Step #5: Step #4: ----> 6 train("1012100")
Step #5: Step #4: 7 train("1031100")
Step #5: Step #4: 9 get_ipython().run_line_magic('cd', '..')
Step #5: Step #4:
Step #5: Step #4: Cell In [20], line 30, in train(category)
Step #5: Step #4: 28 # Train the Prophet model
Step #5: Step #4: 29 model = Prophet()
Step #5: Step #4: ---> 30 model.fit(df_forecast)
Step #5: Step #4: 32 # Save the model locally
Step #5: Step #4: 33 joblib.dump(model, "{}.sav".format(category))
Step #5: Step #4:
Step #5: Step #4: File /workspace/workspace/env/lib/python3.9/site-packages/prophet/forecaster.py:1113, in Prophet.fit(self, df, **kwargs)
Step #5: Step #4: 1111 history = df[df['y'].notnull()].copy()
Step #5: Step #4: 1112 if history.shape[0] < 2:
Step #5: Step #4: -> 1113 raise ValueError('Dataframe has less than 2 non-NaN rows.')
Step #5: Step #4: 1114 self.history_dates = pd.to_datetime(pd.Series(df['ds'].unique(), name='ds')).sort_values()
Step #5: Step #4: 1116 history = self.setup_dataframe(history, initialize_scales=True)
Step #5: Step #4:
Step #5: Step #4: ValueError: Dataframe has less than 2 non-NaN rows.