A set of common machine learning tasks solved using ml ops practices.
custom_train_job: trains a model by packing dependencies and testing code locally using docker. Pushes container to cloud, as a Vertex AI custom job, to take advantage of GPUs.
hf_pipeline: a model training pipeline using Kubeflow and Vertex AI pipelines. Registers model metrics and parameters between executions for experiment tracking.
label_studio: deploys a labeling tool with Terraform based on this article. Serves the GUI as a serverless app using Cloud Run. Annotations are saved using Cloud SQL and Cloud Storage. It exposes new variables for app scaling and security.