ODSC-39737 : allow to use GenericModel.predict() locally (#127)

Author: mingkang111

ads/model/generic_model.py

@@ -2497,6 +2497,7 @@ def predict(
         self,
         data: Any = None,
         auto_serialize_data: bool = False,
+        local: bool = False,
         **kwargs,
     ) -> Dict[str, Any]:
         """Returns prediction of input data run against the model deployment endpoint.
@@ -2521,6 +2522,8 @@ def predict(
             Whether to auto serialize input data. Defauls to `False` for GenericModel, and `True` for other frameworks.
             `data` required to be json serializable if `auto_serialize_data=False`.
             If `auto_serialize_data` set to True, data will be serialized before sending to model deployment endpoint.
+        local: bool.
+            Whether to invoke the prediction locally. Default to False.
         kwargs:
             content_type: str, used to indicate the media type of the resource.
             image: PIL.Image Object or uri for the image.
@@ -2539,10 +2542,21 @@ def predict(
         NotActiveDeploymentError
             If model deployment process was not started or not finished yet.
         ValueError
-            If `data` is empty or not JSON serializable.
+            If model is not deployed yet or the endpoint information is not available.
         """
-        if not self.model_deployment:
-            raise ValueError("Use `deploy()` method to start model deployment.")
+        if local:
+            return self.verify(
+                data=data, auto_serialize_data=auto_serialize_data, **kwargs
+            )
+
+        if not (self.model_deployment and self.model_deployment.url):
+            raise ValueError(
+                "Error invoking the remote endpoint as the model is not "
+                "deployed yet or the endpoint information is not available. "
+                "Use `deploy()` method to start model deployment. "
+                "If you intend to invoke inference using locally available "
+                "model artifact, set parameter `local=True`"
+            )
 
         current_state = self.model_deployment.state.name.upper()
         if current_state != ModelDeploymentState.ACTIVE.name:

tests/unitary/with_extras/model/test_generic_model.py

@@ -170,6 +170,20 @@
 
 INFERENCE_CONDA_ENV = "oci://bucket@namespace/<path_to_service_pack>"
 TRAINING_CONDA_ENV = "oci://bucket@namespace/<path_to_service_pack>"
+DEFAULT_PYTHON_VERSION = "3.8"
+MODEL_FILE_NAME = "fake_model_name"
+FAKE_MD_URL = "http://<model-deployment-url>"
+
+
+def _prepare(model):
+    model.prepare(
+        inference_conda_env=INFERENCE_CONDA_ENV,
+        inference_python_version=DEFAULT_PYTHON_VERSION,
+        training_conda_env=TRAINING_CONDA_ENV,
+        training_python_version=DEFAULT_PYTHON_VERSION,
+        model_file_name=MODEL_FILE_NAME,
+        force_overwrite=True,
+    )
 
 
 class TestEstimator:
@@ -315,14 +329,7 @@ def test_prepare_with_custom_scorepy(self, mock_signer):
     @patch("ads.common.auth.default_signer")
     def test_verify_without_reload(self, mock_signer):
         """Test verify input data without reload artifacts."""
-        self.generic_model.prepare(
-            inference_conda_env="oci://service-conda-packs@ociodscdev/service_pack/cpu/General_Machine_Learning_for_CPUs/1.0/mlcpuv1",
-            inference_python_version="3.6",
-            training_conda_env="oci://service-conda-packs@ociodscdev/service_pack/cpu/Oracle_Database_for_CPU_Python_3.7/1.0/database_p37_cpu_v1",
-            training_python_version="3.7",
-            model_file_name="fake_model_name",
-            force_overwrite=True,
-        )
+        _prepare(self.generic_model)
         self.generic_model.verify(self.X_test.tolist())
 
         with patch("ads.model.artifact.ModelArtifact.reload") as mock_reload:
@@ -332,20 +339,10 @@ def test_verify_without_reload(self, mock_signer):
     @patch("ads.common.auth.default_signer")
     def test_verify(self, mock_signer):
         """Test verify input data"""
-        self.generic_model.prepare(
-            inference_conda_env="oci://service-conda-packs@ociodscdev/service_pack/cpu/General_Machine_Learning_for_CPUs/1.0/mlcpuv1",
-            inference_python_version="3.6",
-            training_conda_env="oci://service-conda-packs@ociodscdev/service_pack/cpu/Oracle_Database_for_CPU_Python_3.7/1.0/database_p37_cpu_v1",
-            training_python_version="3.7",
-            model_file_name="fake_model_name",
-            force_overwrite=True,
-        )
+        _prepare(self.generic_model)
         prediction_1 = self.generic_model.verify(self.X_test.tolist())
         assert isinstance(prediction_1, dict), "Failed to verify json payload."
 
-        prediction_2 = self.generic_model.verify(self.X_test.tolist())
-        assert isinstance(prediction_2, dict), "Failed to verify input data."
-
     def test_reload(self):
         """test the reload."""
         pass
@@ -637,11 +634,31 @@ def test_deploy_with_default_display_name(self, mock_deploy):
             == random_name[:-9]
         )
 
+    @pytest.mark.parametrize("input_data", [(X_test.tolist())])
+    @patch("ads.common.auth.default_signer")
+    def test_predict_locally(self, mock_signer, input_data):
+        _prepare(self.generic_model)
+        test_result = self.generic_model.predict(data=input_data, local=True)
+        expected_result = self.generic_model.estimator.predict(input_data).tolist()
+        assert (
+            test_result["prediction"] == expected_result
+        ), "Failed to verify input data."
+
+        with patch("ads.model.artifact.ModelArtifact.reload") as mock_reload:
+            self.generic_model.predict(
+                data=input_data, local=True, reload_artifacts=False
+            )
+            mock_reload.assert_not_called()
+
     @patch.object(ModelDeployment, "predict")
     @patch("ads.common.auth.default_signer")
     @patch("ads.common.oci_client.OCIClientFactory")
+    @patch(
+        "ads.model.deployment.model_deployment.ModelDeployment.url",
+        return_value=FAKE_MD_URL,
+    )
     def test_predict_with_not_active_deployment_fail(
-        self, mock_client, mock_signer, mock_predict
+        self, mock_url, mock_client, mock_signer, mock_predict
     ):
         """Ensures predict model fails in case of model deployment is not in an active state."""
         with pytest.raises(NotActiveDeploymentError):
@@ -661,7 +678,11 @@ def test_predict_with_not_active_deployment_fail(
 
     @patch("ads.common.auth.default_signer")
     @patch("ads.common.oci_client.OCIClientFactory")
-    def test_predict_bytes_success(self, mock_client, mock_signer):
+    @patch(
+        "ads.model.deployment.model_deployment.ModelDeployment.url",
+        return_value=FAKE_MD_URL,
+    )
+    def test_predict_bytes_success(self, mock_url, mock_client, mock_signer):
         """Ensures predict model passes with bytes input."""
         with patch.object(
             ModelDeployment, "state", new_callable=PropertyMock
@@ -670,7 +691,7 @@ def test_predict_bytes_success(self, mock_client, mock_signer):
             with patch.object(ModelDeployment, "predict") as mock_predict:
                 mock_predict.return_value = {"result": "result"}
                 self.generic_model.model_deployment = ModelDeployment(
-                    model_deployment_id="test"
+                    model_deployment_id="test",
                 )
                 # self.generic_model.model_deployment.current_state = ModelDeploymentState.ACTIVE
                 self.generic_model._as_onnx = False
@@ -683,7 +704,11 @@ def test_predict_bytes_success(self, mock_client, mock_signer):
 
     @patch("ads.common.auth.default_signer")
     @patch("ads.common.oci_client.OCIClientFactory")
-    def test_predict_success(self, mock_client, mock_signer):
+    @patch(
+        "ads.model.deployment.model_deployment.ModelDeployment.url",
+        return_value=FAKE_MD_URL,
+    )
+    def test_predict_success(self, mock_url, mock_client, mock_signer):
         """Ensures predict model passes with valid input parameters."""
         with patch.object(
             ModelDeployment, "state", new_callable=PropertyMock
@@ -800,7 +825,11 @@ def test_from_model_artifact(
 
     @patch("ads.common.auth.default_signer")
     @patch("ads.common.oci_client.OCIClientFactory")
-    def test_predict_success__serialize_input(self, mock_client, mock_signer):
+    @patch(
+        "ads.model.deployment.model_deployment.ModelDeployment.url",
+        return_value=FAKE_MD_URL,
+    )
+    def test_predict_success__serialize_input(self, mock_url, mock_client, mock_signer):
         """Ensures predict model passes with valid input parameters."""
 
         df = pd.DataFrame([1, 2, 3])