How can I get/export a production model from a trained model (after refit) with autosklearn?

[seraus]

Hello team,
I'm doing some tests with autosklearn and I would like to know I can get/export a trained autosklearn model for production use.
Could you please explain which is the process I have to follow?
Currently I am doing this:

automl = autosklearn.classification.AutoSklearnClassifier(
    time_left_for_this_task=3600,
    per_run_time_limit=1800,
    n_jobs=2,
    tmp_folder='/workspace/automl/unito/tmp',
    delete_tmp_folder_after_terminate=False,
    memory_limit=15240,
    resampling_strategy='cv',
    resampling_strategy_arguments={'folds':10},
    metric=autosklearn.metrics.f1_macro,
    scoring_functions=[
        autosklearn.metrics.accuracy,
        autosklearn.metrics.precision,
        autosklearn.metrics.recall,
        autosklearn.metrics.f1_macro
    ]
    include={'classifier': ['random_forestt']},
)
(...)

automl.fit(X_train, y_train, X_test, y_test, dataset_name='mydataset')

y_hat = automl.predict(X_test)
print("Accuracy score", sklearn.metrics.accuracy_score(y_test, y_hat))
print("F1 score", sklearn.metrics.f1_score(y_test, y_hat, average='macro'))
    

automl.refit(X, Y)
y_hat=automl.predict(X_test)
print("F1 score", sklearn.metrics.f1_score(y_test, y_hat, average='macro'))
print("Accuracy score", sklearn.metrics.accuracy_score(y_test, y_hat))

And then I'm using "automl "as a sklearn model and I'm logging it on mlflow but It is rather big (>1 GB).
Is this the right way or there are better ways?

For now I don't need an automatic way, a manual process is sufficient for my tests

Thanks a lot
Sergio

Short Question Description

A clear single sentence question we can try to help with?

With some extra context to follow it up. This way the question is clear for both you and us without it being lost in the paragraph.
Some useful information to help us with your question:

• How did this question come about?
• Would a small code snippet help?
• What have you already looked at?

Before you ask, please have a look at the

• Documentation
  • If it's related but not clear, please include it in your question with a link, we'll try to make it better!
• Examples
  • Likewise, an example can answer many questions! However we can't cover all question with examples but if you think your question would benefit from an example, let us know!
• Issues
  • We try to label all questions with the label Question, maybe someone has already asked. If the question is about a feature, try searching more of the issues. If you find something related but doesn't directly answer your question, please link to it with #(issue number)!

System Details (if relevant)

• Which version of auto-sklearn are you using?

• Are you running this on Linux / Mac / ... ?

• auto-sklearn==0.14.6

• linux centos 7

[seraus]

Hello, is there any news on this issue? Let me know if the question is unclear...
Thanks
Sergio

[eddiebergman]

Hi @seraus,

Many apologies, this slipped under the radar.

So currently we do not have any nice sklearn only export, i.e. there will always be a little auto-sklearn in there. It is something we want to get done eventually but we are currently quite busy with quite a few non-autosklearn things. I hope to be able to return to it in August when the conference period is over.

In the meantime, I recommend checking out automl.show_models() to get access to the models but a more comprehensive answer can be found here. It's not pretty and it's on v0.16 list of things to do.

We are always happy to receive contributions as well :)

Best,
Eddie

[seraus]

Thanks for the answers, @eddiebergman
I'll check the code you suggested
Regards
Sergio