The Deepchecks integration enables you to seamlessly incorporate comprehensive data integrity, data drift, model drift and model performance checks into your ZenML pipelines. By leveraging Deepchecks' extensive library of pre-configured tests, you can easily validate the quality and reliability of the datasets and models used in your ML workflows, ensuring more robust and production-ready pipelines.
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.base import ClassifierMixin
from sklearn.ensemble import RandomForestClassifier
from zenml import pipeline, step
from zenml.integrations.constants import DEEPCHECKS, SKLEARN
from deepchecks.tabular.datasets.classification import iris
from typing_extensions import Tuple, Annotated
from zenml.artifacts.artifact_config import ArtifactConfig
LABEL_COL = "target"
@step
def data_loader() -> Tuple[
Annotated[
pd.DataFrame, ArtifactConfig(name="reference_dataset")
],
Annotated[
pd.DataFrame,
ArtifactConfig(name="comparison_dataset"),
],
]:
"""Load the iris dataset."""
iris_df = iris.load_data(data_format="Dataframe", as_train_test=False)
df_train, df_test = train_test_split(
iris_df, stratify=iris_df[LABEL_COL], random_state=0
)
return df_train, df_test
@step
def trainer(df_train: pd.DataFrame) -> Annotated[ClassifierMixin, ArtifactConfig(name="model")]:
# Train Model
rf_clf = RandomForestClassifier(random_state=0)
rf_clf.fit(df_train.drop(LABEL_COL, axis=1), df_train[LABEL_COL])
return rf_clf
from zenml.integrations.deepchecks.steps import (
deepchecks_data_integrity_check_step,
)
data_validator = deepchecks_data_integrity_check_step.with_options(
parameters=dict(
dataset_kwargs=dict(label=LABEL_COL, cat_features=[]),
),
)
from zenml.integrations.deepchecks.steps import (
deepchecks_data_drift_check_step,
)
data_drift_detector = deepchecks_data_drift_check_step.with_options(
parameters=dict(dataset_kwargs=dict(label=LABEL_COL, cat_features=[]))
)
from zenml.integrations.deepchecks.steps import (
deepchecks_model_validation_check_step,
)
model_validator = deepchecks_model_validation_check_step.with_options(
parameters=dict(
dataset_kwargs=dict(label=LABEL_COL, cat_features=[]),
),
)
from zenml.integrations.deepchecks.steps import (
deepchecks_model_drift_check_step,
)
model_drift_detector = deepchecks_model_drift_check_step.with_options(
parameters=dict(
dataset_kwargs=dict(label=LABEL_COL, cat_features=[]),
),
)
from zenml.config import DockerSettings
docker_settings = DockerSettings(required_integrations=[DEEPCHECKS, SKLEARN])
@pipeline(enable_cache=True, settings={"docker": docker_settings})
def data_validation_pipeline():
"""Links all the steps together in a pipeline"""
df_train, df_test = data_loader()
data_validator(dataset=df_train)
data_drift_detector(
reference_dataset=df_train,
target_dataset=df_test,
)
model = trainer(df_train)
model_validator(dataset=df_train, model=model)
model_drift_detector(
reference_dataset=df_train, target_dataset=df_test, model=model
)
if __name__ == "__main__":
# Run the pipeline
data_validation_pipeline()
In the code above, Deepchecks is integrated with ZenML to perform various data validation and model validation checks. The data_loader step loads the Iris dataset and splits it into training and testing sets. The trainer step trains a RandomForestClassifier on the training data. The deepchecks_data_integrity_check_step is used to validate the integrity of the training data, while the deepchecks_data_drift_check_step detects any data drift between the training and testing datasets. The deepchecks_model_validation_check_step validates the trained model on the training data, and the deepchecks_model_drift_check_step checks for model drift between the training and testing datasets. These steps are linked together in a ZenML pipeline, which is configured to use Docker settings for the required integrations. This setup ensures comprehensive data and model validation using Deepchecks within a ZenML pipeline.
.png)
