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

XGBoost expecting different number of features than custom model #137

Open
hspekt opened this issue Dec 17, 2024 · 6 comments
Open

XGBoost expecting different number of features than custom model #137

hspekt opened this issue Dec 17, 2024 · 6 comments
Labels

Comments

@hspekt
Copy link

hspekt commented Dec 17, 2024

Describe the bug
Using a custom model trained by MQ > Tools > MS/MS intensity prediction errors on library prediction in MaxDIA.

The error message seems to indicate XGBoost is looking for 227 features but only finding 216.

To Reproduce
Steps to reproduce the behavior using MQ 2.6.6.0:

-Create a model using Tools > MS/MS intensity prediction (I used the msms_short.txt provided in a previously closed issue)
-Specify this custom model to be used by MaxDIA > Library type > Predicted > Custom
-Error on Library_prediction_0

Error
... Check failed: learner_model_param_.num_feature >= p_fmat->Info().num_col_ (216 vs. 227) : Number of columns does not match number of features in booster._ ...

Many thanks for any ideas or help with this! Are there parameters I can specify for building with the expected 227 features?

@comaecliptic
Copy link
Collaborator

Hi! Can you attach mqpar.xml file of this run?

@hspekt
Copy link
Author

hspekt commented Dec 18, 2024

mqpar.xml.zip

Cheers thanks!

@hspekt
Copy link
Author

hspekt commented Dec 19, 2024

I have an update to my error description:

The correct error message I receive when doing the described process is:

Prediction model not available for fragmentation type: HCD_ at MsLib.Predict.FragPredictModelCollection.GetModel(FragmentationType fragType, String[] varMods) in C:\Users\bi\source\repos\net7\net\MsLib\Predict\FragPredictModelCollection.cs:line 61__

However, I can see the metadata file in the custom model folder describes it as HCD.

The error message in my first post occurred when I replaced the files in bin/conf/fragModel/standard/unmodified/ with my custom model.

@comaecliptic
Copy link
Collaborator

Yes, specifying a custom model folder only works if the models lie inside the same folder structure as the default ones (so you need to have standard/hcd/unmodified inside the folder you set in MQ).
Okay, I don't see anything wrong in mqpar. Did you change any parameters when training the models?

@hspekt
Copy link
Author

hspekt commented Dec 20, 2024

No change to the parameters when training the models, although I did try with and without the default variable modifications.

When I supply the standard/hcd/unmodified file structure I get a different error:

L:\HS\xxx\combined\proc Library_prediction_0 0 Library_prediction_0 (01/16) Process 77 0 L:\HS\xxx\mqpar.xml 0 0 16_Never get here._ at MsLib.Fragment.FragmentationTypes.FromName(String fragName)

@comaecliptic
Copy link
Collaborator

Can you try specifying standard as Custom directory? It is working for me.
I'm running now using the same MQ version and the same msms to generate library but still unable to reproduce the issue.
Did you copy the metadata file as well when copying model files to the default directory?

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

No branches or pull requests

2 participants