fix: readme docs

 - add example handler
 - add example yamls

kubernetes/kfserving/handler/v2/README.md

@@ -0,0 +1,5 @@
+# Handler for KFServing v2 protocol
+
+## For KFServing v2 protocol the tensor output should be converted to list of list with `[output.flatten().tolist()]` from the postprocess method in the handler.
+
+### Refer: [mnist handler](./mnist_handler.py) and [bert_handler](./bert_handler.py)

kubernetes/kfserving/examples/v2/mnist_handler.py → kubernetes/kfserving/handler/v2/mnist_handler.py

@@ -26,4 +26,3 @@ def postprocess(self, data):
             list : A list of dictionary with predictons and explanations are returned.
         """
         return [data.argmax(1).flatten().tolist()]
-

kubernetes/kfserving/reference_yaml/torchserve-deployment/v1/ts-sample-gpu.yaml

@@ -0,0 +1,13 @@
+apiVersion: "serving.kubeflow.org/v1beta1"
+kind: "InferenceService"
+metadata:
+ name: "torch-pred-gpu"
+spec:
+ predictor:
+   pytorch:
+     protocolVersion: v1
+     storageUri: "pvc://model-store-claim"
+     resources:
+       limits:
+         memory: 4Gi
+         nvidia.com/gpu: 1

kubernetes/kfserving/reference_yaml/torchserve-deployment/v1/ts_sample-gs.yaml

@@ -0,0 +1,9 @@
+apiVersion: "serving.kubeflow.org/v1beta1"
+kind: "InferenceService"
+metadata:
+ name: "torch-pred-gs"
+spec:
+ predictor:
+   pytorch:
+     protocolVersion: v1
+     storageUri: "gs://kfserving-examples/models/torchserve/image_classifier"

kubernetes/kfserving/reference_yaml/torchserve-deployment/v1/ts_sample.yaml

@@ -0,0 +1,9 @@
+apiVersion: "serving.kubeflow.org/v1beta1"
+kind: "InferenceService"
+metadata:
+ name: "torch-pred"
+spec:
+ predictor:
+   pytorch:
+     protocolVersion: v1
+     storageUri: "pvc://model-store-claim"