Dockerize the entier wrapping process for sklearn example

- use multi-stage build and Alpine as a base
- use stripped-down model image
- provide Makefile for running docker build command
- declare all Python dependencies in one requirements.txt file
- copy sklearn_iris_deployment.json

examples/models/sklearn_iris_docker/.dockerignore

@@ -0,0 +1,4 @@
+.dockerignore
+Dockerfile
+Makefile
+sklearn_iris_deployment.json

examples/models/sklearn_iris_docker/Dockerfile

@@ -0,0 +1,84 @@
+## Use alpine as build time and runtime image
+FROM alpine:3.7 as build-alpine
+
+## Install build dependencies
+RUN apk add --update \
+    build-base \
+    freetype-dev \
+    gcc \
+    gfortran \
+    libc6-compat \
+    libffi-dev \
+    libpng-dev \
+    openblas-dev \
+    openssl-dev \
+    py2-pip \
+    python2 \
+    python2-dev\
+    wget \
+    && true
+
+## Symlink missing header, so we can compile numpy
+RUN ln -s /usr/include/locale.h /usr/include/xlocale.h
+
+## Copy package manager config to staging root tree
+RUN mkdir -p /out/etc/apk && cp -r /etc/apk/* /out/etc/apk/
+## Install runtime dependencies under staging root tree
+RUN apk add --no-cache --initdb --root /out \
+    alpine-baselayout \
+    busybox \
+    ca-certificates \
+    freetype \
+    libc6-compat \
+    libffi \
+    libpng \
+    libstdc++ \
+    musl \
+    openblas \
+    openssl \
+    python2 \
+    && true
+## Remove package manager residuals
+RUN rm -rf /out/etc/apk /out/lib/apk /out/var/cache
+
+## Enter model source tree and install all Python depenendcies
+COPY . /src
+WORKDIR /src
+## TODO this does take a while to build, maybe a good idea to
+## put all related build dependencies into a separate public image
+RUN pip install --requirement requirements.txt
+## Train the model
+RUN python train_iris.py
+
+## Copy source code and Python dependencies to the saging root tree
+RUN mkdir -p /out/src && cp -r /src/* /out/src/
+RUN mkdir -p /out/usr/lib/python2.7/ && cp -r /usr/lib/python2.7/* /out/usr/lib/python2.7/
+
+## Use Seldon Core wrapper image to wrap the model source code
+FROM seldonio/core-python-wrapper:0.4 as build-wrapper
+
+ARG MODEL_NAME
+ARG IMAGE_VERSION
+ARG IMAGE_REPO
+
+## Copy staging diretory here
+COPY --from=build-alpine /out /out
+## Wrap the Python model
+WORKDIR /wrappers/python
+RUN python wrap_model.py /out/src $MODEL_NAME $IMAGE_VERSION $IMAGE_REPO --force
+
+## Copy wrapped model source code into staging tree and cleanup what is not neccessary at runtime
+RUN mkdir -p /out/microservice && cp -r /out/src/build/* /out/microservice/ && rm -rf /out/src
+WORKDIR /out/microservice
+RUN rm -f Dockerfile Makefile requirements*.txt build_image.sh push_image.sh
+## TODO dockerfile doesn't support build argument interpolation in array notation for ENTRYPOINT & CMD
+## to get rid of `/bin/sh` wrapper, it'd help to make $MODEL_NAME an environment variable and let the
+## Python script pick it up
+RUN printf '#!/bin/sh\nexec python microservice.py %s REST --service-type MODEL --persistence 0' $MODEL_NAME > microservice.sh && chmod +x microservice.sh
+
+## Copy staging root tree onto an empty image
+FROM scratch
+COPY --from=build-wrapper /out /
+WORKDIR /microservice
+EXPOSE 5000
+ENTRYPOINT ["/microservice/microservice.sh"]

examples/models/sklearn_iris_docker/IrisClassifier.py

@@ -0,0 +1,9 @@
+from sklearn.externals import joblib
+
+class IrisClassifier(object):
+
+    def __init__(self):
+        self.model = joblib.load('IrisClassifier.sav')
+
+    def predict(self,X,features_names):
+        return self.model.predict_proba(X)

examples/models/sklearn_iris_docker/Makefile

@@ -0,0 +1,11 @@
+IMAGE_REPO?=seldonio
+IMAGE_NAME?=irisclassifier
+IMAGE_VERSION?=0.1
+MODEL_NAME?=IrisClassifier
+
+container_image:
+	docker build \
+	  --build-arg IMAGE_REPO=$(IMAGE_REPO) \
+	  --build-arg IMAGE_VERSION=$(IMAGE_VERSION) \
+	  --build-arg MODEL_NAME=$(MODEL_NAME) \
+	  --tag $(IMAGE_REPO)/$(IMAGE_NAME):$(IMAGE_VERSION) .

examples/models/sklearn_iris_docker/requirements.txt

@@ -0,0 +1,9 @@
+numpy==1.11.2
+pandas==0.18.1
+grpc==0.3.post19
+grpcio==1.8.4
+Flask==0.11.1
+futures
+redis==2.10.5
+scipy==0.18.1
+scikit-learn==0.19.0

examples/models/sklearn_iris_docker/sklearn_iris_deployment.json

@@ -0,0 +1,53 @@
+{
+    "apiVersion": "machinelearning.seldon.io/v1alpha1",
+    "kind": "SeldonDeployment",
+    "metadata": {
+        "labels": {
+            "app": "seldon"
+        },
+        "name": "seldon-deployment-example"
+    },
+    "spec": {
+        "annotations": {
+            "project_name": "Iris classification",
+            "deployment_version": "0.1"
+        },
+        "name": "sklearn-iris-deployment",
+        "oauth_key": "oauth-key",
+        "oauth_secret": "oauth-secret",
+        "predictors": [
+            {
+                "componentSpec": {
+                    "spec": {
+                        "containers": [
+                            {
+                                "image": "seldonio/irisclassifier:0.1",
+                                "imagePullPolicy": "IfNotPresent",
+                                "name": "sklearn-iris-classifier",
+                                "resources": {
+                                    "requests": {
+                                        "memory": "1Mi"
+                                    }
+                                }
+                            }
+                        ],
+                        "terminationGracePeriodSeconds": 20
+                    }
+                },
+                "graph": {
+                    "children": [],
+                    "name": "sklearn-iris-classifier",
+                    "endpoint": {
+                        "type" : "REST"
+                    },
+                    "type": "MODEL"
+                },
+                "name": "sklearn-iris-predictor",
+                "replicas": 1,
+                "annotations": {
+                "predictor_version" : "0.1"
+                }
+            }
+        ]
+    }
+}

examples/models/sklearn_iris_docker/train_iris.py

@@ -0,0 +1,25 @@
+import numpy as np
+import os
+from sklearn.linear_model import LogisticRegression
+from sklearn.pipeline import Pipeline
+from sklearn.externals import joblib
+from sklearn import datasets
+
+def main():
+    clf = LogisticRegression()
+    p = Pipeline([('clf', clf)])
+    print 'Training model...'
+    p.fit(X, y)
+    print 'Model trained!'
+
+    filename_p = 'IrisClassifier.sav'
+    print 'Saving model in %s' % filename_p
+    joblib.dump(p, filename_p)
+    print 'Model saved!'
+
+if __name__ == "__main__":
+    print 'Loading iris data set...'
+    iris = datasets.load_iris()
+    X, y = iris.data, iris.target
+    print 'Dataset loaded!'
+    main()