Commands Reference

Here's a commands reference page for Vertex AI, formatted as markdown tables and code blocks:

Vertex AI Commands Reference

Vertex AI CLI Commands

Command	Description
gcloud ai-platform models list	List all models in your project
gcloud ai-platform models create <model-name> --regions=<region>	Create a new model
gcloud ai-platform versions create <version-name> --model=<model-name> --origin=<path-to-model>	Create a new version for a model
gcloud ai-platform predict --model=<model-name> --version=<version-name> --json-instances=<path-to-json-file>	Make a prediction using a model
gcloud ai-platform jobs submit training <job-name> --package-uris=<path-to-package> --region=<region> --python-version=<python-version> --runtime-version=<runtime-version> --job-dir=<job-dir>	Submit a training job
gcloud ai-platform jobs list	List all training jobs in your project
gcloud ai-platform datasets list	List all datasets in your project
gcloud ai-platform datasets create <dataset-name> --description=<description> --metadata-schema-uri=<schema-uri>	Create a new dataset
gcloud ai-platform datasets import --dataset=<dataset-name> --source-uri=<gcs-path>	Import data into a dataset

Vertex AI Python SDK

from google.cloud import aiplatform

# Initialize the Vertex AI client
aiplatform.init(project="<your-project-id>", location="<region>")

# Create a dataset
dataset = aiplatform.Dataset.create(
    display_name="<dataset-name>",
    metadata_schema_uri="<schema-uri>",
    gcs_source=["<gcs-path>"]
)

# Create a model
model = aiplatform.Model.upload(
    display_name="<model-name>",
    artifact_uri="<path-to-model-artifacts>",
    serving_container_image_uri="<container-image>"
)

# Deploy a model
endpoint = model.deploy(
    machine_type="<machine-type>",
    min_replica_count=1,
    max_replica_count=3
)

# Make a prediction
response = endpoint.predict(instances=[{"key": "value"}])

Useful Snippets

Create a custom container image for model deployment

# Dockerfile
FROM tensorflow/serving:latest-gpu

COPY model/ /models/1

Create a custom training container image

# Dockerfile
FROM gcr.io/cloud-aiplatform/training/tf-cpu.2-3:latest

COPY training_script.py /app/
ENTRYPOINT ["python", "/app/training_script.py"]

Use Cloud Storage for model artifacts and training data

# Upload model artifacts to GCS
gcs_path = "gs://<your-bucket>/<path-to-model>"
model.upload(artifact_uri=gcs_path)

# Use GCS for training data
dataset.gcs_source = ["gs://<your-bucket>/<path-to-data>"]

Monitor training jobs using the Vertex AI Dashboard

1. Go to the Vertex AI Dashboard
2. Select the training job you want to monitor
3. View the job details, logs, and metrics

This reference covers the most common Vertex AI commands, SDK usage, and useful snippets. Remember to replace the placeholders (<...>) with your specific values.