A custom container is a Docker image that you create to run your training application. By running your machine learning (ML) training job in a custom container, you can use ML frameworks, non-ML dependencies, libraries, and binaries that are not otherwise supported on Vertex AI.
How training with containers works
Your training application, implemented in the ML framework of your choice, is the core of the training process.
Create an application that trains your model, using the ML framework of your choice.
Decide whether to use a custom container. There could be a pre-built container that already supports your dependencies. Otherwise, you need to build a custom container for your training job. In your custom container, you pre-install your training application and all its dependencies onto an image that is used to run your training job.
Store your training and verification data in a source that Vertex AI can access. To simplify authentication and reduce latency, store your data in Cloud Storage, Cloud Bigtable, or another Google Cloud storage service in the same Google Cloud project and region that you are using for Vertex AI. Learn more about the ways Vertex AI can load your data.
When your application is ready to run, you must build your Docker image and push it to Artifact Registry, Container Registry, or Docker Hub, making sure that Vertex AI can access your registry.
Vertex AI sets up resources for your job. It allocates one or more virtual machines (called training instances) based on your job configuration. You set up a training instance by using the custom container you specify as part of the
WorkerPoolSpecobject when you submit your custom training job.
Vertex AI runs your Docker image, passing through any command-line arguments you specify when you create the training job.
When your training job succeeds or encounters an unrecoverable error, Vertex AI halts all job processes and cleans up the resources.
Advantages of custom containers
Custom containers let you specify and pre-install all the dependencies needed for your application.
- Faster start-up time. If you use a custom container with your dependencies pre-installed, you can save the time that your training application would otherwise take to install dependencies when starting up.
- Use the ML framework of your choice. If you can't find an Vertex AI pre-built container with the ML framework you want to use, you can build a custom container with your chosen framework and use it to run jobs on Vertex AI. For example, you can use a customer container to train with PyTorch.
- Extended support for distributed training. With custom containers, you can do distributed training using any ML framework.
- Use the newest version. You can also use the latest build or minor version
of an ML framework. For example, you can
build a custom container to train with
Hyperparameter tuning with custom containers
To do hyperparameter tuning on Vertex AI, you specify goal metrics, along with whether to minimize or maximize each metric. For example, you might want to maximize your model accuracy, or minimize your model loss. You also list the hyperparameters you'd like to tune, along with the range of acceptable values for each hyperparameter. Vertex AI does multiple trials of your training application, tracking and adjusting the hyperparameters after each trial. When the hyperparameter tuning job is complete, Vertex AI reports values for the most effective configuration of your hyperparameters, and a summary for each trial.
To do hyperparameter tuning with custom containers, you need to make the following adjustments:
- In your Dockerfile: install
- In your training code:
See how to configure a hyperparameter tuning job that uses custom containers or learn more about how hyperparameter tuning works on Vertex AI.
GPUs in custom containers
For training with GPUs, your custom container needs to meet a few special requirements. You must build a different Docker image than what you'd use for training with CPUs.
- Pre-install the CUDA toolkit and cuDNN in your Docker image. The recommended
way to build a custom container with support for GPUs is to use the
nvidia/cudaimage as your base image for your custom container. The
nvidia/cudacontainer image has matching versions of CUDA toolkit and cuDNN pre-installed, and it helps you set up the related environment variables correctly.
- Install your training application, along with your required ML framework and other dependencies in your Docker image.
- Learn more about how to create a custom container for your training job.