fix: Add fit_full_data param to skip validation for full-dataset training.#1421
fix: Add fit_full_data param to skip validation for full-dataset training.#1421murunlin wants to merge 5 commits intomicrosoft:mainfrom
Conversation
|
Hi,
Sorry for lack of clarity. The title of PR: fix:mrl0409 #1421
Thank you for the feedback!
Best regards,
Mu Runlin
穆润琳
北京邮电大学/研博生/人工智能学院
北京
------------------ Original ------------------
From: ***@***.***>;
Date: Wed, Apr 9, 2025 05:46 PM
To: ***@***.***>;
Cc: ***@***.***>; ***@***.***>;
Subject: Re: [microsoft/FLAML] fix:mrl0409 (PR #1421)
@thinkall commented on this pull request.
Hi @murunlin , thank you very much for the PR. Could you please elaborate the description/title of the PR? You can refer to #1405 , #1385 as examples.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you were mentioned.Message ID: ***@***.***>
|
|
@microsoft-github-policy-service agree
|
|
@microsoft-github-policy-service agree |
|
@thinkall Hi Li, thanks a lot for your guidance! I have updated the description and checked other parts. Please let me know if there is anything missed. |
@thinkall I had some misunderstanding that the dataset split should be explicitly implemented. —— I'll promptly look into another issue and submit a new PR ASAP. Appreciate your time! |

Why are these changes needed?
These changes address the issue that fitting the model using all available data (#1404).
This modification set the retrain_full parameter to True, which could handle the split_ratio=0 case in resampling stage, enabling to retrain model on the full dataset.
Related issue number
Related to #1404
Checks