Generating models with CoreML in Sonoma slowed significantly

I have accidentally updated to Sonoma, and found my CoreML models are generating nearly 7x slower since the update. I also no longer get the verbose information in terminal (i.e time taken per cycle, deviation from actual result etc). This is using xcode and swift developed for MacOS.

The M1 laptop I am using is also under considerably less stress (i.e it is no longer getting warm)

Is there a flag I need to set to increase performance, a button I need to press, any suggestions would be helpful.

Please note this is 24hr+ since the update, so it should no longer be affected by any usual background tasks after an upgrade.

Post not yet marked as solved Up vote post of frankthechicken Down vote post of frankthechicken
626 views
Posted by
frankthechicken
Reply

  • Just to note, I am using this to build the model

    let params_boosted = MLBoostedTreeRegressor.ModelParameters(maxIterations: 10000) let model = try MLBoostedTreeRegressor(trainingData: trainingdata, targetColumn: columntopredict, parameters: params_boosted)

    on an M1 MacBook Pro 16Gb. Models are literally taking 7x longer to build.

    —
    frankthechicken
Add a Comment

Replies

Have you tried the same with Sonoma developer beta build and see if it's still the issue there?

Posted by
spacebrother
Post not yet marked as solved Up vote reply of spacebrother Down vote reply of spacebrother
Add a Comment

I have not yet, I didn't really want to update to Sonoma and only did so accidentally!!

I am willing to try if there is the possibility that these issues might be fixed. It's either that or downgrading back to Ventura.

Posted by
frankthechicken
Up vote reply of frankthechicken Down vote reply of frankthechicken
Add a Comment