Updated pr.

Author: lu-ohai

ads/model/deployment/model_deployment.py

@@ -29,6 +29,11 @@
 from ads.model.common.utils import _is_json_serializable
 from ads.model.deployment.common.utils import send_request
 from ads.model.deployment.model_deployment_infrastructure import (
+    DEFAULT_BANDWIDTH_MBPS,
+    DEFAULT_REPLICA,
+    DEFAULT_SHAPE_NAME,
+    DEFAULT_OCPUS,
+    DEFAULT_MEMORY_IN_GBS,
     MODEL_DEPLOYMENT_INFRASTRUCTURE_TYPE,
     ModelDeploymentInfrastructure,
 )
@@ -64,12 +69,6 @@
 MODEL_DEPLOYMENT_TYPE = "modelDeployment"
 MODEL_DEPLOYMENT_INFERENCE_SERVER_TRITON = "TRITON"
 
-MODEL_DEPLOYMENT_INSTANCE_SHAPE = "VM.Standard.E4.Flex"
-MODEL_DEPLOYMENT_INSTANCE_OCPUS = 1
-MODEL_DEPLOYMENT_INSTANCE_MEMORY_IN_GBS = 16
-MODEL_DEPLOYMENT_INSTANCE_COUNT = 1
-MODEL_DEPLOYMENT_BANDWIDTH_MBPS = 10
-
 MODEL_DEPLOYMENT_RUNTIMES = {
     ModelDeploymentRuntimeType.CONDA: ModelDeploymentCondaRuntime,
     ModelDeploymentRuntimeType.CONTAINER: ModelDeploymentContainerRuntime,
@@ -1601,7 +1600,7 @@ def _build_model_deployment_configuration_details(self) -> Dict:
 
         instance_configuration = {
             infrastructure.CONST_INSTANCE_SHAPE_NAME: infrastructure.shape_name
-            or MODEL_DEPLOYMENT_INSTANCE_SHAPE,
+            or DEFAULT_SHAPE_NAME,
         }
 
         if instance_configuration[infrastructure.CONST_INSTANCE_SHAPE_NAME].endswith(
@@ -1613,14 +1612,14 @@ def _build_model_deployment_configuration_details(self) -> Dict:
                 infrastructure.CONST_OCPUS: infrastructure.shape_config_details.get(
                     "ocpus", None
                 )
-                or MODEL_DEPLOYMENT_INSTANCE_OCPUS,
+                or DEFAULT_OCPUS,
                 infrastructure.CONST_MEMORY_IN_GBS: infrastructure.shape_config_details.get(
                     "memory_in_gbs", None
                 )
                 or infrastructure.shape_config_details.get(
                     "memoryInGBs", None
                 )
-                or MODEL_DEPLOYMENT_INSTANCE_MEMORY_IN_GBS,
+                or DEFAULT_MEMORY_IN_GBS,
             }
 
         if infrastructure.subnet_id:
@@ -1629,7 +1628,7 @@ def _build_model_deployment_configuration_details(self) -> Dict:
         scaling_policy = {
             infrastructure.CONST_POLICY_TYPE: "FIXED_SIZE",
             infrastructure.CONST_INSTANCE_COUNT: infrastructure.replica
-            or MODEL_DEPLOYMENT_INSTANCE_COUNT,
+            or DEFAULT_REPLICA,
         }
 
         if not runtime.model_uri:
@@ -1660,7 +1659,7 @@ def _build_model_deployment_configuration_details(self) -> Dict:
 
         model_configuration_details = {
             infrastructure.CONST_BANDWIDTH_MBPS: infrastructure.bandwidth_mbps
-            or MODEL_DEPLOYMENT_BANDWIDTH_MBPS,
+            or DEFAULT_BANDWIDTH_MBPS,
             infrastructure.CONST_INSTANCE_CONFIG: instance_configuration,
             runtime.CONST_MODEL_ID: model_id,
             infrastructure.CONST_SCALING_POLICY: scaling_policy,

ads/model/deployment/model_deployment_infrastructure.py

@@ -22,7 +22,9 @@
 DEFAULT_BANDWIDTH_MBPS = 10
 DEFAULT_WEB_CONCURRENCY = 10
 DEFAULT_REPLICA = 1
-DEFAULT_SHAPE_NAME = "VM.Standard.E2.4"
+DEFAULT_SHAPE_NAME = "VM.Standard.E4.Flex"
+DEFAULT_OCPUS = 1
+DEFAULT_MEMORY_IN_GBS = 16
 
 logger = logging.getLogger(__name__)
 
@@ -625,4 +627,8 @@ def init(self) -> "ModelDeploymentInfrastructure":
             .with_web_concurrency(self.web_concurrency or DEFAULT_WEB_CONCURRENCY)
             .with_replica(self.replica or DEFAULT_REPLICA)
             .with_shape_name(self.shape_name or DEFAULT_SHAPE_NAME)
+            .with_shape_config_details(
+                ocpus=self.shape_config_details.get(self.CONST_OCPUS, DEFAULT_OCPUS),
+                memory_in_gbs=self.shape_config_details.get(self.CONST_MEMORY_IN_GBS, DEFAULT_MEMORY_IN_GBS)
+            )
         )