ODSC-29065: adding more unit test

Author: Ziqun Ye

ads/model/deployment/model_deployment.py

@@ -863,9 +863,9 @@ def predict(
             and `json_input` required to be json serializable. If `auto_serialize_data` set
             to True, data will be serialized before sending to model deployment endpoint.
         model_name: str
-            Defaults to None. When the `Inference_server="triton"`, the name of the model to invoke.
+            Defaults to None. When the `inference_server="triton"`, the name of the model to invoke.
         model_version: str
-            Defaults to None. When the `Inference_server="triton"`, the version of the model to invoke.
+            Defaults to None. When the `inference_server="triton"`, the version of the model to invoke.
         kwargs:
             content_type: str
                 Used to indicate the media type of the resource.

ads/opctl/backend/ads_model_deployment.py

@@ -126,7 +126,12 @@ def watch(self) -> None:
     def predict(self) -> None:
         ocid = self.config["execution"].get("ocid")
         data = self.config["execution"].get("payload")
+        model_name = self.config["execution"].get("model_name")
+        model_version = self.config["execution"].get("model_version")
         with AuthContext(auth=self.auth_type, profile=self.profile):
             model_deployment = ModelDeployment.from_id(ocid)
-            data = json.loads(data)
-            print(model_deployment.predict(data))
+            try:
+                data = json.loads(data)
+            except:
+                pass
+            print(model_deployment.predict(data=data, model_name=model_name, model_version=model_version))

ads/opctl/cli.py

@@ -558,6 +558,8 @@ def deactivate(**kwargs):
     required=False,
     help="The conda env used to load the model and conduct the prediction. This is only used when model id is passed to `ocid` and a local predict is conducted. It should match the inference conda env specified in the runtime.yaml file which is the conda pack being used when conducting real model deployment.",
 )
+@click.option("--model-version", nargs=1, required=False, help="When the `inference_server='triton'`, the version of the model to invoke. This can only be used when model deployment id is passed in. For the other cases, it will be ignored.")
+@click.option("--model-name", nargs=1, required=False, help="When the `inference_server='triton'`, the name of the model to invoke. This can only be used when model deployment id is passed in. For the other cases, it will be ignored.")
 @click.option("--debug", "-d", help="set debug mode", is_flag=True, default=False)
 def predict(**kwargs):
     """

ads/opctl/script.py

@@ -5,7 +5,7 @@
 from ads.model.generic_model import GenericModel
 
 
-def verify(artifact_dir, data, compartment_id, project_id):
+def verify(artifact_dir, data, compartment_id, project_id): # pragma: no cover
     with tempfile.TemporaryDirectory() as td:
         model = GenericModel.from_model_artifact(
             uri=artifact_dir,
@@ -15,17 +15,20 @@ def verify(artifact_dir, data, compartment_id, project_id):
             project_id=project_id,
         )
 
-        data = json.loads(data)
+        try:
+            data = json.loads(data)
+        except:
+            pass
         print(model.verify(data, auto_serialize_data=False))
 
 
-def main():
+def main(): # pragma: no cover
     args = sys.argv[1:]
     verify(
         artifact_dir=args[0], data=args[1], compartment_id=args[2], project_id=args[3]
     )
     return 0
 
 
-if __name__ == "__main__":
+if __name__ == "__main__": # pragma: no cover
     main()

tests/unitary/with_extras/opctl/test_opctl_model_deployment_backend.py

@@ -20,13 +20,17 @@ def config(self):
                 "oci_config": "~/.oci/config",
                 "oci_profile": "DEFAULT",
                 "run_id": "test_model_deployment_id",
+                "ocid": "fake_model_id",
                 "auth": "api_key",
                 "wait_for_completion": False,
                 "max_wait_time": 1000,
                 "poll_interval": 12,
                 "log_type": "predict",
                 "log_filter": "test_filter",
                 "interval": 3,
+                "payload": "fake_payload",
+                "model_name": "model_name",
+                "model_version": "model_version",
             }
         }
 
@@ -103,3 +107,15 @@ def test_watch(self, mock_from_id, mock_watch):
         mock_watch.assert_called_with(
             log_type="predict", interval=3, log_filter="test_filter"
         )
+
+    @patch("ads.opctl.backend.ads_model_deployment.ModelDeployment.predict")
+    @patch("ads.opctl.backend.ads_model_deployment.ModelDeployment.from_id")
+    def test_predict(self, mock_from_id, mock_predict):
+        config = self.config
+        mock_from_id.return_value = ModelDeployment()
+        backend = ModelDeploymentBackend(config)
+        backend.predict()
+        mock_from_id.assert_called_with("fake_model_id")
+        mock_predict.assert_called_with(
+            data="fake_payload", model_name='model_name', model_version='model_version'
+        )