black_friday_challenge_deploy.sh

PROJECT_ID=<PROJECT_ID>
BUCKET_NAME="${PROJECT_ID}-black-friday-demo-bucket"
MODEL_NAME=black_friday_forecast
REGION=us-central1
VERSION='v'`date '+%Y%m%d_%H%M%S'`
MODEL_PATH=gs://$BUCKET_NAME/models/$VERSION/model
PACKAGE_PATH=gs://$BUCKET_NAME/models/$VERSION/package/black_friday_forecast-0.0.1.tar.gz

# Package solution
python setup.py sdist

# Upload code to GCS
gsutil cp ./dist/black_friday_forecast-0.0.1.tar.gz $PACKAGE_PATH

# Upload model to GCS
gsutil cp ./models/model.joblib $MODEL_PATH/model.joblib

# Create model if it doesn't exist
gcloud ai-platform models create $MODEL_NAME --regions $REGION --project $PROJECT_ID

# Create a version
gcloud --quiet beta ai-platform versions create $VERSION \
  --model $MODEL_NAME \
  --origin $MODEL_PATH \
  --runtime-version 1.15 \
  --python-version 2.7 \
  --framework SCIKIT_LEARN \
  --package-uris $PACKAGE_PATH \
  --project $PROJECT_ID