Create a model using custom training

Developing a custom training application and creating a model using custom training.

Getting started

• Training code requirements

  Describes requirements to consider as you write training code.

• Containerizing and running code locally

  Describes how to use the gcloud CLI to containerize and run training code on your local computer.

• Understanding the custom training service on Vertex AI

  Describes the lifecycle of a training cluster during a distributed training job, and explains how the Vertex AI custom training service handles errors.

• Using managed datasets

  Describes how to use Vertex AI managed datasets with custom training.

• Configuring container settings for custom training

  Describes the fields of the Vertex AI API that you must specify to configure training container settings for either a custom container or a Python training application that runs on a pre-built container.

• Configuring compute resources for custom training

  Describes the different compute resources that you can use for custom training and how to configure them.

Pre-built containers

• Creating a Python training application for a pre-built container

  How to create a Python source distribution that contains your training application and upload it to a Cloud Storage bucket.

• Pre-built containers for custom training

  Provides a list of the pre-built containers for training, and describes how to use them with a Python training application.

Custom containers

• Custom containers overview

  An overview of custom containers - a Docker image that you create to run your training application.

• Creating a custom container image for training

  How to build a custom container image to perform custom training.

Custom training methods

• Choosing a custom training method.

  Provides an overview and comparison of the different ways you can run custom training.

• Creating custom training jobs

  How to create custom training jobs to run your custom training applications on Vertex AI.

• Creating training pipelines

  How to create training pipelines to run custom training applications on Vertex AI

Distributed training

• Using distributed training

  How to run distributed training jobs on Vertex AI.

Hyperparameter tuning

• Overview of hyperparameter tuning

  Describes the concepts involved in hyperparameter tuning, which is the automated model enhancer provided by Vertex AI.

• Using hyperparameter tuning

  How to prepare your training application for hyperparameter tuning, and create your hyperparameter training job.

Export model artifacts

• Exporting model artifacts for prediction

  Requirements for saving a model in the form of one or more model artifacts so Vertex AI can serve predictions.

Monitor training

• Monitoring and debugging training with an interactive shell

  Learn how to use an interactive shell to inspect your training container while it runs.

Training in a private network

• Using private IP with custom training

  Learn how set up a custom training job that is peered with a private network.