update handler; package dependencies

Author: kukushking

modules/sagemaker/sagemaker-model-monitoring/sagemaker_model_monitoring/baselining_construct.py

@@ -47,7 +47,12 @@ def __init__(
                     command=[
                         "bash",
                         "-c",
-                        "pip install -r requirements.txt -t /asset-output && cp -au . /asset-output",
+                        "pip install pydantic==2.10.3 -t /asset-output --no-cache-dir && "
+                        "pip install sagemaker==2.232.2 -t /asset-output --no-deps --no-cache-dir && "
+                        "rm -rf /asset-output/{boto*,urllib3*,certifi*,six*,python_dateutil*,jmespath*,s3transfer*} && "
+                        "find /asset-output \\( -name '*.pyc' -o -name '*.so' \\) -delete && "
+                        "find /asset-output -type d \\( -name '__pycache__' -o -name 'test*' \\) -exec rm -rf {} + && "
+                        "cp -au . /asset-output",
                     ],
                 ),
             ),
@@ -87,8 +92,8 @@ def __init__(
                     "s3:PutObject",
                 ],
                 resources=[
-                    f"{baseline_training_data_s3_uri}*",
-                    f"{baseline_output_data_s3_uri}*",
+                    f"arn:aws:s3:::{baseline_training_data_s3_uri.replace('s3://', '')}*",
+                    f"arn:aws:s3:::{baseline_output_data_s3_uri.replace('s3://', '')}*",
                 ],
             )
         )

modules/sagemaker/sagemaker-model-monitoring/sagemaker_model_monitoring/lambda/baselining_handler.py

@@ -2,7 +2,6 @@
 from typing import Any, Dict
 
 import boto3
-import pandas as pd
 from models import DataQualityParams, ModelBiasParams, ModelExplainabilityParams, ModelQualityParams
 from sagemaker.clarify import BiasConfig, DataConfig, ModelConfig, ModelPredictedLabelConfig, SHAPConfig
 from sagemaker.model_monitor import (
@@ -151,12 +150,9 @@ def start_baselining_job(event: Dict[str, Any]) -> Dict[str, Any]:
             dataset_type=params.dataset_type,
         )
 
-        # Use mean value of test dataset as SHAP baseline
-        test_dataframe = pd.read_csv(training_data_uri, header=None)
-        shap_baseline = [list(test_dataframe.mean())]
-
+        # Use configurable SHAP baseline
         shap_config = SHAPConfig(
-            baseline=shap_baseline,
+            baseline=params.shap_baseline,
             num_samples=params.num_samples,
             agg_method=params.agg_method,
             save_local_shap_values=params.save_local_shap_values,

modules/sagemaker/sagemaker-model-monitoring/sagemaker_model_monitoring/lambda/models.py

@@ -49,6 +49,7 @@ class ModelExplainabilityParams(BaseParams):
     num_samples: int = Field(default=100, ge=1, description="Number of samples for SHAP")
     agg_method: str = Field(default="mean_abs", description="SHAP aggregation method")
     save_local_shap_values: bool = Field(default=False, description="Save local SHAP values")
+    shap_baseline: List[List[float]] = Field(default=[[0.0]], description="SHAP baseline values")
 
     def get_headers(self) -> Optional[List[str]]:
         return [h.strip() for h in self.headers.split(",")] if self.headers else None