CatBoost¶

The Snowflake ML Model Registry supports models created using CatBoost (models derived from catboost.CatBoost, such as catboost.CatBoostClassifier, catboost.CatBoostRegressor, and catboost.CatBoostRanker).

The following additional options can be used in the options dictionary when you call log_model:

Option	Description
`target_methods`	A list of the names of the methods available on the model object. CatBoost models have the following target methods by default, assuming the method exists: `predict`, `predict_proba`.
`enable_explainability`	Whether to enable explainability for the model using SHAP. Defaults to `True`. When enabled, an `explain` method will be available on the logged model.
`cuda_version`	The version of the CUDA runtime to be used when deploying to a platform with GPU; defaults to 11.8. If manually set to `None`, the model cannot be deployed to a platform having a GPU.

You must specify either the sample_input_data or signatures parameter when logging a CatBoost model so that the registry knows the signatures of the target methods.

Examples¶

These examples assume reg is an instance of snowflake.ml.registry.Registry.

CatBoostClassifier¶

The following example demonstrates the key steps to train a CatBoost classifier, log it to the Snowflake ML Model Registry, and use the registered model for inference and explainability. The workflow includes:

Trains a CatBoost classifier on a sample dataset.
Logs the model to the Snowflake ML Model Registry.
Makes predictions and retrieves prediction probabilities.
Gets SHAP values for the model’s predictions.

import catboost
from sklearn import datasets, model_selection

# Load dataset
cal_data = datasets.load_breast_cancer(as_frame=True)
cal_X = cal_data.data
cal_y = cal_data.target

# Normalize column names (replace spaces with underscores)
cal_X.columns = [col.replace(' ', '_') for col in cal_X.columns]

cal_X_train, cal_X_test, cal_y_train, cal_y_test = model_selection.train_test_split(
    cal_X, cal_y, test_size=0.2
)

# Train CatBoost Classifier
classifier = catboost.CatBoostClassifier(
    iterations=100,
    learning_rate=0.1,
    depth=6,
    verbose=False
)
classifier.fit(cal_X_train, cal_y_train)

# Log the model
model_ref = reg.log_model(
    model=classifier,
    model_name="my_catboost_classifier",
    version_name="v1",
    sample_input_data=cal_X_test,
)

# Make predictions
result_df = model_ref.run(cal_X_test[-10:], function_name="predict")

# Get prediction probabilities
proba_df = model_ref.run(cal_X_test[-10:], function_name="predict_proba")

# Get explanations (SHAP values)
explanations_df = model_ref.run(cal_X_test[-10:], function_name="explain")

Copy

CatBoostRegressor¶

The following example demonstrates the key steps to train a CatBoost regressor, log it to the Snowflake ML Model Registry, and use the registered model for inference. The workflow includes:

Trains a CatBoost regressor on a sample dataset.
Logs the model to the Snowflake ML Model Registry.
Makes predictions.

import catboost
from sklearn import datasets, model_selection

# Load dataset
cal_data = datasets.load_diabetes(as_frame=True)
cal_X = cal_data.data
cal_y = cal_data.target

cal_X_train, cal_X_test, cal_y_train, cal_y_test = model_selection.train_test_split(
    cal_X, cal_y, test_size=0.2
)

# Train CatBoost Regressor
regressor = catboost.CatBoostRegressor(
    iterations=100,
    learning_rate=0.1,
    depth=6,
    verbose=False
)
regressor.fit(cal_X_train, cal_y_train)

# Log the model
model_ref = reg.log_model(
    model=regressor,
    model_name="my_catboost_regressor",
    version_name="v1",
    sample_input_data=cal_X_test,
)

# Make predictions
result_df = model_ref.run(cal_X_test[-10:], function_name="predict")

Copy

Disabling Explainability¶

If you do not need explainability features, you can disable them during logging to reduce model size and dependencies:

model_ref = reg.log_model(
    model=classifier,
    model_name="my_catboost_classifier_no_explain",
    version_name="v1",
    sample_input_data=cal_X_test,
    options={"enable_explainability": False},
)

Copy