Set Up and Manage an AWS EC2 Cluster

GenAI Studio can be installed on a cluster that is hosted on an AWS EC2 instance.

Prerequisites #

Before you start, you must meet the following prerequisites.

Machine Learning Development Environment Prerequisites

• Install Docker.
• Install Pip3.

AWS Account Prerequisites

• An AWS account with IAM credentials.
• Install the AWS CLI configure it.
• Gather your EC2 key pair name or create one.
• Your key pair name should be short and without capital letters.

Supported GPUs

GenAI Studio supports the following instance types:

• A100 - 40GB (p4d.24xlarge)
• V100 - 16GB (p3.16xlarge)
• T4 - (g4dn.12xlarge)

See also: Supported hardware reference.

Set Up det deploy #

To access the Determined core det deploy CLI tool, you may choose the official versions or latest developer versions.

Official Versions #

• Set up det deploy using official versions.
• The official versions include instructions on how to Deploy a cluster on Amazon Web Services (AWS).

Latest Developer Versions #

To set up det deploy using the latest developer versions, first install the developer version of the CLI tool and then use it to deploy to AWS. For additional information, you can visit installation from source.

• Clone the determined-ai/core repository.

 git clone https://github.com/determined-ai/determined.git

• Go to the main branch

cd determined && git checkout main

• Install the determined package from the source.

cd harness && pip install .

Additional Arguments #

Apart from the official det deploy instructions and args, consider these additional arguments for deploying GenAI Studio:

• deployment-type genai: This is used for deploying GenAI Studio services and configurations.
• genai-version (optional): This option allows you to choose a specific version by replacing latest with one of the image tags available in the DockerHub repository.
• compute-agent-instance-type (optional): This option enables you to choose a specific instance type. Refer to the Supported GPUs section on this page for more information.

Verify the Installation #

To verify the installation, navigate to GenAI Studio:

<ec2_public_domain>:8080/genai