Fix model serving part of xgboost_ams_housing example. (#478)

* Fix model serving for ames house example

* change the step instructions

* add public image

xgboost_ames_housing/README.md

@@ -93,30 +93,18 @@ The model is exported to the location `/tmp/ames/housing.dat`. We will use [Seld
  * `requirements.txt`
 
 ## Model Serving Locally
-We are going to use [seldon-core](https://github.com/SeldonIO/seldon-core/) to serve the model. [HousingServe.py](seldon_serve/HousingServe.py) contains the code to serve the model. Run the following command to create a microservice 
+We are going to use [seldon-core](https://github.com/SeldonIO/seldon-core/) to serve the model. [HoussingServe.py](seldon_serve/HousingServe.py) contains the code to serve the model. You can find seldon core model wrapping details [here](https://github.com/SeldonIO/seldon-core/blob/master/docs/wrappers/python.md). The seldon-core microservice image can be built by the following command.
 
 ```
-docker run -v $(pwd):/seldon_serve seldonio/core-python-wrapper:0.7 /seldon_serve HousingServe 0.1 gcr.io --base-image=python:3.6 --image-name=${PROJECT_ID}/housingserve
+cd seldon_serve && s2i build . seldonio/seldon-core-s2i-python2:0.4 gcr.io/${PROJECT_ID}/housingserve:latest --loglevel=3
 ```
 
-Let's build the seldon-core microservice image. You can find seldon core model wrapping details [here](https://github.com/SeldonIO/seldon-core/blob/master/docs/wrappers/python.md).
+
+
+Let's run the docker image locally.
 
 ```
-cd build
+docker run -p 5000:5000 gcr.io/${PROJECT_ID}/housingserve:latest
-./build_image.sh
-```
-
-You should see the docker image locally `gcr.io/cloudmlplat/housingserve` which can be run locally to serve the model. Before running the image locally push it to `gcr.io`
-
-```
-gcloud auth configure-docker
-docker push gcr.io/${PROJECT_ID}/housingserve:0.1
-```
-
-Let's run the docker image now
-
-```
-docker run -p 5000:5000 gcr.io/cloudmlplat/housingserve:0.1
 ```
 
 Now you are ready to send requests on `localhost:5000`
@@ -125,6 +113,7 @@ Now you are ready to send requests on `localhost:5000`
 curl -H "Content-Type: application/x-www-form-urlencoded" -d 'json={"data":{"tensor":{"shape":[1,37],"values":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37]}}}' http://localhost:5000/predict
 ```
 
+The response looks like this.
 ```
 {
   "data": {
@@ -163,12 +152,19 @@ tf-hub-0                                  1/1       Running   0          16m
 tf-job-dashboard-7b57c549c8-bfpp8         1/1       Running   0          16m
 tf-job-operator-594d8c7ddd-lqn8r          1/1       Running   0          16m
 ```
-Deploy the XGBoost model
+
+Second, we need to upload our previously built docker image to `gcr.io`. A public image is available at `gcr.io/kubeflow-examples/housingserve:latest`
 
 ```
-ks generate seldon-serve-simple xgboost-ames   \
+gcloud auth configure-docker
+docker push gcr.io/${PROJECT_ID}/housingserve:latest
+```
+
+Finally, we can deploy the XGBoost model
+```
+ks generate seldon-serve-simple-v1alpha2 xgboost-ames   \
                                 --name=xgboost-ames   \
-                                --image=gcr.io/cloudmlplat/housingserve:0.1   \
+                                --image=gcr.io/${PROJECT_ID}/housingserve:latest   \
                                 --namespace=${NAMESPACE}   \
                                 --replicas=1
                                 

xgboost_ames_housing/seldon_serve/.s2i/environment

@@ -0,0 +1,4 @@
+MODEL_NAME=HousingServe
+API_TYPE=REST
+SERVICE_TYPE=MODEL
+PERSISTENCE=0