Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add feedback for forecasting #1258

Merged
merged 27 commits into from
Nov 16, 2023
Merged

Add feedback for forecasting #1258

merged 27 commits into from
Nov 16, 2023

Conversation

americast
Copy link
Member

@americast americast commented Oct 5, 2023

Provide feedback when Forecasting UDF is called in the following ways:

  • Reporting confidence intervals
  • Returning a metric for the forecasting performance.
  • Providing suggestions in simple special cases, such as during Flat predictions.

Eg:

SELECT HomeForecast();
SUGGESTION: Predictions are flat. Consider using LIBRARY 'neuralforecast' for more accrate predictions.

Partially fixes #1257 and #1243.

@americast americast requested a review from xzdandy October 5, 2023 20:30
@americast americast self-assigned this Oct 5, 2023
@americast
Copy link
Member Author

Another concern with this approach is that the suggestion is being provided through a print statement. Hence, it would appear on the server side, and the user might miss it. Any hints on how we can send the suggestion to the user?

@xzdandy
Copy link
Collaborator

xzdandy commented Oct 21, 2023

Another concern with this approach is that the suggestion is being provided through a print statement. Hence, it would appear on the server side, and the user might miss it. Any hints on how we can send the suggestion to the user?

Good point. I think one option is to make the suggestion as part of the CREATE FUNCTION 's output.

@xzdandy
Copy link
Collaborator

xzdandy commented Oct 24, 2023

Hi @americast, thanks for adding for suggestions a metric for the forecasting performance. Do you think we can add that information into evadb/catalog/models/function_catalog.py or evadb/catalog/models/function_metadata_catalog.py? We can decide which catalog table is better. The idea is to be able to query this information afterwards.

Beside, we can also output this information as part of the output of CREATE FUNCTION, so user can directly view them.

@americast americast force-pushed the feedback branch 2 times, most recently from d382361 to 1a3a338 Compare October 24, 2023 19:17
@americast americast force-pushed the feedback branch 3 times, most recently from 2c28b81 to 77188f0 Compare October 31, 2023 03:55
@americast americast marked this pull request as ready for review October 31, 2023 06:36
@americast americast force-pushed the feedback branch 2 times, most recently from 235070b to c94f1f6 Compare November 11, 2023 18:04
@americast americast force-pushed the feedback branch 2 times, most recently from e03c4d0 to 2eb51b2 Compare November 11, 2023 20:45
@americast
Copy link
Member Author

@xzdandy The issue seems to be coming up due to multithreading while training neuralforecast model on the GPU -- the entire test suite seems to be getting spawned into multiple threads and thus it appears to have got hanged. For now I have removed test for neuralforecast.

@xzdandy xzdandy merged commit 69b39b8 into staging Nov 16, 2023
7 checks passed
@xzdandy xzdandy deleted the feedback branch November 16, 2023 17:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Feedback for the user when using the FORECASTING UDF
2 participants