View a markdown version of this page

AcceleratorConfig - AWS Serverless Application Model

AcceleratorConfig

AcceleratorConfig is a property of AWS::Serverless::Function that configures accelerator resources, such as GPUs, for an AWS Lambda (Lambda) function that runs on Lambda Managed Instances. You specify the amount of accelerator memory that Lambda allocates to each execution environment of the function.

To use AcceleratorConfig, the function must also specify CapacityProviderConfig. Lambda supports accelerator configuration only for functions that use the Image package type and the x86_64 architecture.

Syntax

To declare this entity in your AWS Serverless Application Model (AWS SAM) template, use the following syntax.

YAML

AcceleratorMemorySize: Integer

Properties

AcceleratorConfig has the following property.

AcceleratorMemorySize

The amount of accelerator memory, in GiB, that Lambda allocates to each execution environment.

Valid values: 3, 6, 12, 16, 24, or 48

Type: Integer

Required: Yes

CloudFormation compatibility: This property is passed directly to the AcceleratorMemorySize property of the AWS::Lambda::Function AcceleratorConfig data type.

Examples

GPU function on a capacity provider

The following example defines a container image function that requests 24 GiB of accelerator memory and runs on a capacity provider.

YAML

Resources: MyGpuFunction: Type: AWS::Serverless::Function Properties: PackageType: Image ImageUri: 111122223333.dkr.ecr.us-east-1.amazonaws.com/my-gpu-image:latest Architectures: - x86_64 MemorySize: 2048 CapacityProviderConfig: Arn: !GetAtt MyCapacityProvider.Arn AcceleratorConfig: AcceleratorMemorySize: 24