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

Actual ntrees value in tree based models is incorrect when using train/valid for early stopping #7132

Closed
exalate-issue-sync bot opened this issue May 11, 2023 · 3 comments
Assignees

Comments

@exalate-issue-sync
Copy link

There's a bug where if you set nfolds = 0, and turn on early stopping in any tree based model (GBM, DRF, XGBoost), the reported actual value for the ntrees parameters are incorrect. This is supposed to be updated with the final/found/actual value, but for some reason it’s just replacing it with the “input” value.

When you have CV turned on (and use CV for early stopping), it will correctly report the "actual" (found) value for ntrees (which can be confirmed by looking at model summary with actual # of trees). However, if you turn off CV, then it incorrectly reports the input value as the “actual” value. Bug can be reproduced from the attached notebook.

This is only currently a bug in Python right now because there's a [https://h2oai.atlassian.net/browse/PUBDEV-8051|https://h2oai.atlassian.net/browse/PUBDEV-8051|smart-link] where we are not storing the actual value at all, so that's currently unaffected (but we need to fix that ASAP as well).

@h2o-ops-ro
Copy link
Collaborator

JIRA Issue Details

Jira Issue: PUBDEV-8531
Assignee: Adam Valenta
Reporter: Erin LeDell
State: Resolved
Fix Version: 3.40.0.1
Attachments: Available (Count: 5)
Development PRs: Available

@h2o-ops-ro
Copy link
Collaborator

Attachments From Jira

Attachment Name: H2O-3_AutoML_Starter.html
Attached By: Erin LeDell
File Link:https://h2o-3-jira-github-migration.s3.amazonaws.com/PUBDEV-8531/H2O-3_AutoML_Starter.html

Attachment Name: H2O-3_AutoML_Starter.ipynb
Attached By: Erin LeDell
File Link:https://h2o-3-jira-github-migration.s3.amazonaws.com/PUBDEV-8531/H2O-3_AutoML_Starter.ipynb

Attachment Name: Screen Shot 2022-01-17 at 6.53.37 PM.png
Attached By: Erin LeDell
File Link:https://h2o-3-jira-github-migration.s3.amazonaws.com/PUBDEV-8531/Screen Shot 2022-01-17 at 6.53.37 PM.png

Attachment Name: Screen Shot 2022-01-17 at 6.53.48 PM.png
Attached By: Erin LeDell
File Link:https://h2o-3-jira-github-migration.s3.amazonaws.com/PUBDEV-8531/Screen Shot 2022-01-17 at 6.53.48 PM.png

Attachment Name: Screen Shot 2022-01-17 at 7.08.41 PM.png
Attached By: Erin LeDell
File Link:https://h2o-3-jira-github-migration.s3.amazonaws.com/PUBDEV-8531/Screen Shot 2022-01-17 at 7.08.41 PM.png

@h2o-ops-ro
Copy link
Collaborator

Linked PRs from JIRA

#6476
#6493
#6660

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants