[Python] Snowflake ML: `There is no method with name explain available in the model`

Issue:


I'm working with snowflake-ml-python and trying to use the explain function on an XGBoost model deployed in Snowflake. However, I get the following error when calling run with function_name="explain":

ValueError: There is no method with name explain available in the model ....

Context:

1. I trained an XGBClassifier model using snowflake.ml.modeling.xgboost.using their documentation:
2. Registered it in Snowflake’s model registry with enable_explainability=True.
3. The model successfully trains and makes predictions.
4. When calling model_version.run(features_df, function_name="explain"), I get the error above.

Code Snippet:


Environment:

• windows 10
• python 3.8
• snowflake-ml-python==1.5.4
• snowflake-snowpark-python==1.19.0

Model Training and Registration:


import pandas as pd
from sklearn.datasets import make_classification
from snowflake.ml.modeling.xgboost import XGBClassifier
from snowflake.snowpark import Session
from snowflake.ml.registry import registry

# Snowflake session setup
conn_params = {
"user": "<...>",
"account": "<...>",
"warehouse": "<...>",
"database": "<...>",
"schema": "<...>",
"role": "<...>",
"authenticator": "externalbrowser",
}

session = Session.builder.configs(conn_params).create()

# Define columns
FEATURE_COLS = ["X1", "X2", "X3", "X4", "X5", "X6"]
LABEL_COLS = ["Y"]
OUTPUT_COLS = ["PREDICTIONS"]

# Generate data
X, y = make_classification(n_samples=40000, n_features=6, n_informative=4, n_redundant=1, random_state=0, shuffle=True)
X_df = pd.DataFrame(X, columns=FEATURE_COLS)
y_df = pd.DataFrame(y, columns=LABEL_COLS)
features_pandas = pd.concat([X_df, y_df], axis=1)
features_df = session.create_dataframe(features_pandas)

# Train and register model
xgboost_model = XGBClassifier(input_cols=FEATURE_COLS, label_cols=LABEL_COLS, output_cols=OUTPUT_COLS)
xgboost_model.fit(features_df)

native_registry = registry.Registry(session=session, database_name=session.get_current_database(), schema_name=session.get_current_schema())
model_name = "Model_Only"
version = "V3"

model_version = native_registry.log_model(
model=xgboost_model,
model_name=model_name,
version_name=version,
sample_input_data=features_df.sample(n=1000),
conda_dependencies=["snowflake-ml-python"],
comment="Model_Only test",
options={"enable_explainability": True}
)

Making Predictions & Explanation Attempt:


model_version = native_registry.get_model(model_name).version(version)
predictions = model_version.run(features_df, function_name="predict")
print(predictions.show())

# Attempting to explain
explanations = model_version.run(features_df, function_name="explain") # Fails here

Questions:

1. Does XGBClassifier in Snowflake ML support explain? It seems it does from the doc.
2. Is there any additional setup required to enable explainability?
3. If explain is not supported, what is the recommended alternative for feature importance in Snowflake ML?

Any insights would be greatly appreciated!

Continue reading...