Pipeline in Google AI Platform with Kubeflow

• Steps will be explained in more detail.

• dsl-compile --py pipeline.py --output pipeline.yaml

python setup.py sdist --formats=gztar gsutil cp dist/svd_model-0.1.tar.gz gs://mokarakayamodels/staging gcloud ai-platform models create svdmodel --regions us-central1

python setup.py sdist --formats=gztar gsutil cp dist/lightfm_model-0.1.tar.gz gs://mokarakayamodels/staging gsutil cp lightfm.p gs://mokarakayamodels/staging gcloud ai-platform models create lightfmmodelwithlogs --regions us-central1 --enable-logging
--enable-console-logging

gcloud beta ai-platform versions create v0_0_1
--model lightfmmodelwithlogs
--runtime-version 2.2
--python-version 3.7
--origin gs://mokarakayamodels/staging
--package-uris gs://mokarakayamodels/staging/lightfm_model-0.1.tar.gz
--prediction-class lightfm_predictor.MyPredictor
--enable-logging

• See runtime-version-list for Google AI platform at; https://cloud.google.com/ai-platform/training/docs/runtime-version-list