fix: readme doc

 - add logs

Signed-off-by: jagadeesh <jagadeeshj@ideas2it.com>

kubernetes/kserve/kserve_wrapper/README.md

@@ -59,11 +59,11 @@ sudo  mkdir -p /mnt/models/model-store
 
 For v1 protocol
 
-``export TS_SERVICE_ENVELOPE=kserve`
+`export TS_SERVICE_ENVELOPE=kserve`
 
 For v2 protocol
 
-``export TS_SERVICE_ENVELOPE=kservev2`
+`export TS_SERVICE_ENVELOPE=kservev2`
 
 - Step 10: Move the config.properties to /mnt/models/config/.
   The config.properties file is as below :
@@ -93,6 +93,20 @@ torchserve --start --ts-config /mnt/models/config/config.properties
 
 - Step 12: Run the below command to start the KFServer
 
+- Step 13: Set protocol version
+
+For v1 protocol
+
+`export PROTOCOL_VERSION=v1`
+
+For v2 protocol
+
+`export PROTOCOL_VERSION=v2`
+
+For grpc protocol v2 format set
+
+`export PROTOCOL_VERSION=grpc-v2`
+
 ```bash
 python3 serve/kubernetes/kserve/kserve_wrapper/__main__.py
 ```
@@ -127,7 +141,7 @@ Output:
 
 The curl request for explain is as below:
 
-```
+```bash
 curl -H "Content-Type: application/json" --data @serve/kubernetes/kserve/kf_request_json/v1/mnist.json http://0.0.0.0:8080/v1/models/mnist:explain
 ```
 
@@ -169,7 +183,7 @@ Response:
 
 The curl request for explain is as below:
 
-```
+```bash
 curl -H "Content-Type: application/json" --data @serve/kubernetes/kserve/kf_request_json/v1/mnist.json http://0.0.0.0:8080/v2/models/mnist/explain
 ```
 
@@ -192,6 +206,21 @@ Response:
 }
 ```
 
+For grpc-v2 protocol
+
+- Download the proto file
+
+```bash
+curl -O https://github.com/kserve/kserve/master/docs/predict-api/v2/grpc_predict_v2.proto
+```
+- Download [grpcurl](https://github.com/fullstorydev/grpcurl)
+
+Make gRPC request
+
+```bash
+grpcurl -vv -plaintext -proto grpc_predict_v2.proto -d @ localhost:8081 inference.GRPCInferenceService.ModelInfer <<< $(cat "serve/kubernetes/kserve/kf_request_json/v2/mnist_tensor_bytes_grpc.json")
+```
+
 ## KServe Wrapper Testing in Local for BERT
 
 - Step 1: Follow the same steps from to 10 as what was done for MNIST.

kubernetes/kserve/kserve_wrapper/TorchserveModel.py

@@ -76,6 +76,7 @@ def __init__(
 
         logging.info("Predict URL set to %s", self.predictor_host)
         logging.info("Explain URL set to %s", self.explainer_host)
+        logging.info("Protocol version is %s", self.protocol)
 
     def grpc_client(self):
         if self._grpc_client_stub is None:
@@ -88,6 +89,7 @@ def grpc_client(self):
     async def _grpc_predict(
         self,
         payload: Union[ModelInferRequest, InferRequest],
+        headers: Dict[str, str] = None,
     ) -> ModelInferResponse:
         """Overrides the `_grpc_predict` method in Model class. The predict method calls
         the `_grpc_predict` method if the self.protocol is "grpc_v2"

kubernetes/kserve/kserve_wrapper/__main__.py

@@ -1,6 +1,7 @@
 """ KServe wrapper to handler inference in the kserve_predictor """
 import json
 import logging
+import os
 
 import kserve
 from kserve.model_server import ModelServer
@@ -50,9 +51,6 @@ def parse_config():
     models = keys["model_snapshot"]["models"]
     model_names = []
 
-    protocol = "grpc-v2"
-    # protocol = "v2"
-
     # Get all the model_names
     for model, value in models.items():
         model_names.append(model)
@@ -75,12 +73,11 @@ def parse_config():
         model_store = DEFAULT_MODEL_STORE
 
     logging.info(
-        "Wrapper : Model names %s, inference address %s, management address %s, grpc_inference_address, %s, protocol %s, model store %s",
+        "Wrapper : Model names %s, inference address %s, management address %s, grpc_inference_address, %s, model store %s",
         model_names,
         inference_address,
         management_address,
         grpc_inference_address,
-        protocol,
         model_store,
     )
 
@@ -89,7 +86,6 @@ def parse_config():
         inference_address,
         management_address,
         grpc_inference_address,
-        protocol,
         model_store,
     )
 
@@ -101,10 +97,11 @@ def parse_config():
         inference_address,
         management_address,
         grpc_inference_address,
-        protocol,
         model_dir,
     ) = parse_config()
 
+    protocol = os.environ.get("PROTOCOL_VERSION")
+
     models = []
 
     for model_name in model_names: