add minimal example of deploying pachyderm-trained model

examples/pachyderm/.gitignore

@@ -0,0 +1,2 @@
+data.csv
+model.joblib

examples/pachyderm/deploy.yaml

@@ -0,0 +1,19 @@
+
+apiVersion: machinelearning.seldon.io/v1
+kind: SeldonDeployment
+metadata:
+  name: pachyderm-sklearn
+spec:
+  annotations:
+    seldon.io/executor: "true"
+  name: iris
+  predictors:
+  - componentSpecs:
+    graph:
+      children: []
+      implementation: SKLEARN_SERVER
+      modelUri: s3://pachyderm-iris
+      envSecretRefName: seldon-init-container-secret
+      name: classifier
+    name: default
+    replicas: 1

examples/pachyderm/get-data.py

@@ -0,0 +1,21 @@
+from sklearn import datasets
+import pandas as pd
+import numpy as np
+
+
+def main():
+    print("Getting Iris Dataset")
+    iris = datasets.load_iris()
+    X, y = iris.data, iris.target
+
+    data = pd.DataFrame(
+        data=np.c_[iris["data"], iris["target"]],
+        columns=iris["feature_names"] + ["target"],
+    )
+
+    data.to_csv("data.csv", index=False)
+    print("Iris dataset saved to 'data.csv' file")
+
+
+if __name__ == "__main__":
+    main()

examples/pachyderm/iris-trainer/Dockerfile

@@ -0,0 +1,6 @@
+FROM python:3.7
+
+ADD requirements.txt .
+RUN pip install -r requirements.txt
+
+ADD train_iris.py .

examples/pachyderm/iris-trainer/Makefile

@@ -0,0 +1,14 @@
+VERSION=0.1
+IMAGE_BASE=seldonio/pachyderm-iris-trainer
+
+KIND_NAME ?= kind
+
+
+build:
+	docker build . -t ${IMAGE_BASE}:${VERSION}
+
+push:
+	docker push ${IMAGE_BASE}:${VERSION}
+
+kind_load:
+	kind load -v 3 docker-image ${IMAGE_BASE}:${VERSION} --name ${KIND_NAME}

examples/pachyderm/iris-trainer/requirements.txt

@@ -0,0 +1,5 @@
+scikit-learn == 0.20.3
+numpy >= 1.8.2
+joblib >= 0.13.0
+pandas >= 1.0.1
+pyaml >= 5.3

examples/pachyderm/iris-trainer/train_iris.py

@@ -0,0 +1,33 @@
+import os
+import yaml
+import joblib
+import pandas as pd
+from sklearn.pipeline import Pipeline
+from sklearn.linear_model import LogisticRegression
+
+
+# /pfs/iris-data and /pfs/outs are special directories mounted by Pachyderm
+INPUT_FILE = os.path.join("/pfs/iris-data", "data.csv")
+OUTPUT_FILE = os.path.join("/pfs/out", "model.joblib")
+
+
+def main():
+    clf = LogisticRegression(solver="liblinear", multi_class="ovr")
+    p = Pipeline([("clf", clf)])
+    print("Training model...")
+    p.fit(X, y)
+    print("Model trained!")
+
+    print(f"Saving model in {OUTPUT_FILE}")
+    joblib.dump(p, OUTPUT_FILE)
+    print("Model saved!")
+
+
+if __name__ == "__main__":
+    print(f"Loading iris data set from {INPUT_FILE}")
+    data = pd.read_csv(INPUT_FILE)
+    y = data["target"]
+    X = data[(x for x in data.columns if x != "target")]
+    print("Dataset loaded!")
+
+    main()

examples/pachyderm/metadata.yaml

@@ -0,0 +1,12 @@
+
+name: iris
+versions: [iris/pachyderm:789a6e406b90450185b4280ab16b7f14]
+platform: sklearn
+inputs:
+- datatype: BYTES
+  name: input
+  shape: [ 1, 4 ]
+outputs:
+- datatype: BYTES
+  name: output
+  shape: [ 3 ]

examples/pachyderm/secret.yaml

@@ -0,0 +1,11 @@
+
+apiVersion: v1
+kind: Secret
+metadata:
+  name: seldon-init-container-secret
+type: Opaque
+stringData:
+  AWS_ACCESS_KEY_ID: minioadmin
+  AWS_SECRET_ACCESS_KEY: minioadmin
+  AWS_ENDPOINT_URL: http://minio.minio-system.svc.cluster.local:9000
+  USE_SSL: "false"