Google Cloud ML Examples

Slides and Videos

• [Google Cloud ML Slides](/Google Cloud ML.pdf)
• Google Cloud ML Video
• Google Cloud ML Video (KOREAN)

Simple multiplication (train.1-multiply)

Run locally:

python -m train.1-multiply

Run in ClouldML

Set variables

JOB_NAME=<your job name>
  
JOB_NAME="task8"

PROJECT_ID=`gcloud config list project --format "value(core.project)"`
STAGING_BUCKET=gs://${PROJECT_ID}-ml

Submit a job

gcloud beta ml jobs submit training ${JOB_NAME} \
--package-path=train \
--staging-bucket="${STAGING_BUCKET}" \
--module-name=train.1-multiply

Read input.csv from Google Storage (train.2-input)

Run locally:

python -m train.2-input

Run in ClouldML

Set variables

JOB_NAME="task8"

PROJECT_ID=`gcloud config list project --format "value(core.project)"`
STAGING_BUCKET=gs://${PROJECT_ID}-ml
INPUT_PATH=${STAGING_BUCKET}/input

Copy input.csv to Google Storage

gsutil cp input/input.csv $INPUT_PATH/input.csv

Submit a job

gcloud beta ml jobs submit training ${JOB_NAME} \
--package-path=train \
--staging-bucket="${STAGING_BUCKET}" \
--module-name=train.2-input \
-- --input_dir="${INPUT_PATH}"

Write checkpoint files to Google Storage (train.3-output)

Run locally:

python -m train.3-output

Run in CloudML

Set variables

JOB_NAME="task20"
PROJECT_ID=`gcloud config list project --format "value(core.project)"`
STAGING_BUCKET=gs://${PROJECT_ID}-ml
OUTPUT_PATH=${STAGING_BUCKET}/output/

Create the output folder (Copy an empty file to the GS path with trailing slash, /)

gsutil cp /dev/null $OUTPUT_PATH

Submit a job

gcloud beta ml jobs submit training ${JOB_NAME} \
--package-path=train \
--staging-bucket="${STAGING_BUCKET}" \
--module-name=train.3-output \
-- --output_dir="${OUTPUT_PATH}"

Contributions/Comments

We always welcome your contributions/comments. Use the Issues.