[feature] #1191 is trained property to indicate the state of the model.

[timruhkopf]

#1191

[Copilot]

Pull Request Overview

This PR introduces an is_trained property across multiple model implementations and updates corresponding tests to assert the trained state of the models.

• Adds assertions in tests for RandomForest, GP, and MCMC GP to verify is_trained before and after training
• Modifies the training methods in RandomForest, Multi-objective models, Gaussian process (MCMC based), and AbstractModel to update the _is_trained flag
• Updates the CHANGELOG to document the new trained property

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/test_model/test_rf.py	Added assertions for is_trained in RandomForest tests
tests/test_model/test_gp_mcmc.py	Added assertions for is_trained in GP_MCMC tests
tests/test_model/test_gp.py	Added a dedicated test for is_trained in GP model
smac/model/random_forest/random_forest.py	Sets _is_trained to True after fitting the forest
smac/model/multi_objective_model.py	Implements a read-only is_trained property using all()
smac/model/gaussian_process/mcmc_gaussian_process.py	Changes _is_trained assignment to depend on all model flags
smac/model/abstract_model.py	Initializes _is_trained to False and exposes is_trained
CHANGELOG.md	Documents the new is_trained property feature

[Copilot]

Pull Request Overview

This PR introduces an is_trained property to indicate whether models have been trained. Key changes include:

• Adding tests in models’ test files (rf, gp, gp_mcmc) to verify the state of is_trained before and after training.
• Updating model training methods in random_forest, multi_objective_model, mcmc_gaussian_process, and abstract_model to appropriately set the is_trained flag.
• Documenting the new property in the CHANGELOG.

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tests/test_model/test_rf.py	Added assertions for is_trained before and after training of RandomForest model
tests/test_model/test_gp_mcmc.py	Added assertions for is_trained for the GP MCMC model training
tests/test_model/test_gp.py	Added a new test to validate the is_trained property in the GP model
smac/model/random_forest/random_forest.py	Sets the is_trained flag to True after fitting the random forest model
smac/model/multi_objective_model.py	Introduces is_trained property and sets it based on the training state of internal models
smac/model/gaussian_process/mcmc_gaussian_process.py	Updates is_trained flag setting with a condition based on the presence of internal models
smac/model/abstract_model.py	Initializes the is_trained flag and exposes it via a property
CHANGELOG.md	Updates the changelog to include the new is_trained feature

[benjamc]

Can you make sure that the property is_trained is an abstract property, thus required?
If this is difficult with the property type, you can also create an abstractmethod

[timruhkopf]

Can you make sure that the property is_trained is an abstract property, thus required? If this is difficult with the property type, you can also create an abstractmethod

I have inherited from ABC and made it an abstract property. Now we have a few property getters that will only return if the "_is_trained" attribute is True or False. But future work may want to access that differently, so it kind of makes sense.