Sherif akoush/sc 2543/alibi explain mlserver runtime (#3707)

* Initial POC to integrate mlserver alibi

* Create svc without waiting for deployment to be ready

* add explainer image_v2 in operator configmap.yaml

* map core explainer type in mlserver type

* wire up explainer image_v2

* add  mlserver env constants and use them

* refactor to utils to mlserver.go

* allow to pass explainer init parameters + tests

* add initParameters to explainer in the CRD

* add initial  explainer_examples_v2.ipynb

* use mlserver:0.6.0.dev2

* add integration test for explainer v2 (anchor tabular)

* add  tests

* add v2 explainer doc reference

doc/source/analytics/explainers.md

@@ -39,6 +39,13 @@ For Alibi explainers that need to be trained you should
 
 The runtime environment in our [Alibi Explain Server](https://github.com/SeldonIO/seldon-core/tree/master/components/alibi-explain-server) is locked using [Poetry](https://python-poetry.org/). See our e2e example [here](../examples/iris_explainer_poetry.html) on how to use that definition to train your explainers.
 
+### V2 protocol for explainer using [MLServer](https://github.com/SeldonIO/MLServer) (incubating)
+
+The support for v2 protocol is now handled with MLServer moving forward. This is experimental
+and only works for black-box explainers.
+
+For an e2e example, please check AnchorTabular notebook [here](../examples/iris_anchor_tabular_explainer_v2.html).
+
 ## Explain API
 
 For the Seldon Protocol an endpoint path will be exposed for:
@@ -84,7 +91,7 @@ The explain method is also supported for tensorflow and v2 kfserving protocols.
 | ------ | ----- |
 | seldon | `http://<host>/<ingress-path>/api/v1.0/explain` |
 | tensorflow | `http://<host>/<ingress-path>/v1/models/<model-name>:explain` |
-| kfserving | `http://<host>/<ingress-path>/v2/models/<model-name>/explain` |
+| kfserving | `http://<host>/<ingress-path>/v2/models/<model-name>/infer` |
 
 
 Note: for `tensorflow` protocol we support similar non-standard extension as for the [prediction API](../graph/protocols.md#rest-and-grpc-tensorflow-protocol), `http://<host>/<ingress-path>/v1/models/:explain`.

doc/source/examples/iris_anchor_tabular_explainer_v2.nblink

@@ -0,0 +1,3 @@
+{
+  "path": "../../../notebooks/explainer_examples_v2.ipynb"
+}

helm-charts/seldon-core-operator/templates/customresourcedefinition_seldondeployments.machinelearning.seldon.io.yaml

@@ -4272,6 +4272,8 @@ spec:
                         type: object
                       envSecretRefName:
                         type: string
+                      initParameters:
+                        type: string
                       modelUri:
                         type: string
                       replicas:

helm-charts/seldon-core-operator/templates/customresourcedefinition_v1_seldondeployments.machinelearning.seldon.io.yaml

@@ -5234,6 +5234,8 @@ spec:
                           type: object
                         envSecretRefName:
                           type: string
+                        initParameters:
+                          type: string
                         modelUri:
                           type: string
                         replicas:
@@ -11421,6 +11423,8 @@ spec:
                           type: object
                         envSecretRefName:
                           type: string
+                        initParameters:
+                          type: string
                         modelUri:
                           type: string
                         replicas:
@@ -17608,6 +17612,8 @@ spec:
                           type: object
                         envSecretRefName:
                           type: string
+                        initParameters:
+                          type: string
                         modelUri:
                           type: string
                         replicas:

helm-charts/seldon-core-operator/values.yaml

@@ -27,7 +27,7 @@ certManager:
 
 # ## Install with limited namespace visibility
 # If you want to ensure seldon-core-controller can only have visibility
-#   to specifci namespaces you can set the controllerId
+#   to specify namespaces you can set the controllerId
 controllerId: ""
 
 # Whether operator should create the webhooks and configmap on startup (false means created from chart)
@@ -43,7 +43,7 @@ defaultUserID: "8888"
 managerUserID: 8888
 
 # ## Service Orchestrator (Executor)
-# The executor is the default service orchestrator which has superceeded the "Java Engine"
+# The executor is the default service orchestrator which has superceded the "Java Engine"
 executor:
   port: 8000
   metricsPortName: metrics
@@ -119,15 +119,15 @@ predictor_servers:
         defaultImageVersion: "1.12.0-dev"
         image: seldonio/mlflowserver
       kfserving:
-        defaultImageVersion: "0.5.0"
+        defaultImageVersion: "0.6.0.dev0"
         image: seldonio/mlserver
   SKLEARN_SERVER:
     protocols:
       seldon:
         defaultImageVersion: "1.12.0-dev"
         image: seldonio/sklearnserver
       kfserving:
-        defaultImageVersion: "0.5.0"
+        defaultImageVersion: "0.6.0.dev0"
         image: seldonio/mlserver
   TENSORFLOW_SERVER:
     protocols:
@@ -143,7 +143,7 @@ predictor_servers:
         defaultImageVersion: "1.12.0-dev"
         image: seldonio/xgboostserver
       kfserving:
-        defaultImageVersion: "0.5.0"
+        defaultImageVersion: "0.6.0.dev0"
         image: seldonio/mlserver
   TRITON_SERVER:
     protocols:
@@ -153,7 +153,7 @@ predictor_servers:
   TEMPO_SERVER:
     protocols:
       kfserving:
-        defaultImageVersion: "0.5.0"
+        defaultImageVersion: "0.6.0.dev0"
         image: seldonio/mlserver
 
 # ## Other
@@ -206,3 +206,4 @@ engine:
 # Explainer image
 explainer:
   image: seldonio/alibiexplainer:1.12.0-dev
+  image_v2: seldonio/mlserver:0.6.0.dev2