Develop

examples/README.md

@@ -85,7 +85,8 @@ poetry run mlflow models build-docker --model-uri models:/Diabetes_Prediction_Mo
 
 
 ## Walmart Sales
-This example uses a public dataset on Walmart sales over a specific period. The primary goal of this use case is to demonstrate how to deploy multiple models using a single serving endpoint. Additionally, it highlights some limitations of MLflow's ability to automatically generate the appropriate Dockerfile for model deployment.
+
+This example utilizes a public dataset containing Walmart sales data over a defined period. The main objective is to showcase how to deploy multiple models through a single serving endpoint. In this scenario, a separate model is trained for each store, and all store-specific models are saved as artifacts when registering the final model. This approach enables flexible and scalable deployment for store-level predictions.
 
 * package: `examples/walmart_sales_regression`
 

examples/walmart_sales_regression/base.py

@@ -12,6 +12,8 @@
 from typing import List
 from typing import Optional
 from mlflow.models import infer_signature
+from pathlib import Path
+import platform
 
 
 class WalmartSalesRegressor(mlflow.pyfunc.PythonModel):
@@ -35,13 +37,27 @@ def load_context(self, context):
         :param context: The context object containing the model.
         :return: None
         """
+        if platform.system() == "Linux":
+            # Convert Windows-style paths to POSIX paths for Linux compatibility
+            print(
+                "Converting Windows-style paths to POSIX paths for Linux compatibility..."
+            )
+            context_artifacts = {
+                key: value.replace("\\", "/")
+                for key, value in context.artifacts.items()
+            }
+
+        else:
+            # Use the context artifacts as is for non-Linux systems
+            print("Using context artifacts as is for non-Linux systems...")
+            context_artifacts = context.artifacts
 
-        model_artifacts = context.artifacts
+        print("Loading model artifacts from context...")
         self.models = {
             store_id: mlflow.sklearn.load_model(uri)
-            for store_id, uri in model_artifacts.items()
+            for store_id, uri in context_artifacts.items()
         }
-        print(f"Model artifact URIs loaded: {self.artifact_uris}")
+        print(f"Model artifact URIs loaded: {context_artifacts}")
 
     def fit_model(self, x_train, y_train, store_id: int, run_id: str):
         """

examples/walmart_sales_regression/configs.yaml

@@ -9,6 +9,6 @@ categorical_features:
 
 target: "Weekly_Sales"
 # Model configuration
-registered_model_name: "walmart-store-sales-regressor"
+registered_model_name: "walmart-store-sales-regressor-code"
 run_name: "walmart-sales-regressors"
 artifact_path: "store-sales-regressor"

examples/walmart_sales_regression/inference.py

@@ -28,14 +28,14 @@ def main():
     print("Model loaded.")
 
     # # predicting for the store store_id
-    store_id = "2"
-    x_test = x_test[x_test["Store"] == int(store_id)]
-    y_test = y_test[y_test["Store"] == int(store_id)]
+    store_id = 2
+    x_test = x_test[x_test["Store"] == store_id]
+    y_test = y_test[y_test["Store"] == store_id]
     x_test = x_test.drop(columns=["Store"])
     y_test = y_test.drop(columns=["Store"])
 
     # make predictions
-    predictions = model.predict(x_test, params={"store_id": store_id})
+    predictions = model.predict(x_test, params={"store_id": str(store_id)})
 
     weekly_sales = y_test[configs["target"]].values
     print(

examples/walmart_sales_regression/online_inference.py

@@ -10,7 +10,7 @@ def main():
     Perform online inference using a REST API.
 
     To deploy the model using the local server, run the following command:
-    `poetry run mlflow models serve -m models:/walmart-store-sales-regressor@production -p 5000 --env-manager local`
+    `poetry run mlflow models serve -m models:/walmart-store-sales-regressor-code@production -p 5000 --env-manager local`
     """
 
     url = "http://localhost:5000/invocations"

examples/walmart_sales_regression/train.py

@@ -60,26 +60,8 @@ def main(**kwargs):
             artifacts=store_sales_regressor.artifact_uris,
         )
 
-        # log model with code
-        mlflow.pyfunc.log_model(
-            artifact_path=configs["artifact_path"] + "-code",
-            python_model=store_sales_regressor,
-            infer_code_paths=True,
-            registered_model_name=registered_model_name + "-code",
-            input_example=x_test.sample(5),
-            signature=signature,
-            artifacts=store_sales_regressor.artifact_uris,
-        )
-
-        print("Models fitted successfully.")
-
         set_alias_to_latest_version(
             registered_model_name=registered_model_name,
             alias="production",
             client=kwargs["mlflow_client"],
         )
-        set_alias_to_latest_version(
-            registered_model_name=registered_model_name + "-code",
-            alias="production",
-            client=kwargs["mlflow_client"],
-        )

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "mlflow_for_ml_dev"
-version = "1.5.4"
+version = "1.6.0"
 description = "Code examples for the youtube playlist 'MLflow for Machine Learning Development' by Manuel Gil"
 authors = ["Manuel Gil <[email protected]>"]
 readme = "README.md"