interface AcceleratorCapabilitiesProperty
| Language | Type name |
|---|---|
.NET | Amazon.CDK.AWS.Deadline.CfnFleet.AcceleratorCapabilitiesProperty |
Go | github.com/aws/aws-cdk-go/awscdk/v2/awsdeadline#CfnFleet_AcceleratorCapabilitiesProperty |
Java | software.amazon.awscdk.services.deadline.CfnFleet.AcceleratorCapabilitiesProperty |
Python | aws_cdk.aws_deadline.CfnFleet.AcceleratorCapabilitiesProperty |
TypeScript | aws-cdk-lib » aws_deadline » CfnFleet » AcceleratorCapabilitiesProperty |
Provides information about the GPU accelerators used for jobs processed by a fleet.
Accelerator capabilities cannot be used with wait-and-save fleets. If you specify accelerator capabilities, you must use either spot or on-demand instance market options. > Each accelerator type maps to specific EC2 instance families:
t4: Uses G4dn instance familya10g: Uses G5 instance familyl4: Uses G6 and Gr6 instance familiesl40s: Uses G6e instance family
Example
// The code below shows an example of how to instantiate this type.
// The values are placeholders you should change.
import { aws_deadline as deadline } from 'aws-cdk-lib';
const acceleratorCapabilitiesProperty: deadline.CfnFleet.AcceleratorCapabilitiesProperty = {
selections: [{
name: 'name',
// the properties below are optional
runtime: 'runtime',
}],
// the properties below are optional
count: {
min: 123,
// the properties below are optional
max: 123,
},
};
Properties
| Name | Type | Description |
|---|---|---|
| selections | IResolvable | (IResolvable | Accelerator)[] | A list of accelerator capabilities requested for this fleet. |
| count? | IResolvable | Accelerator | The number of GPU accelerators specified for worker hosts in this fleet. |
selections
Type:
IResolvable | (IResolvable | Accelerator)[]
A list of accelerator capabilities requested for this fleet.
Only Amazon Elastic Compute Cloud instances that provide these capabilities will be used. For example, if you specify both L4 and T4 chips, AWS Deadline Cloud will use Amazon EC2 instances that have either the L4 or the T4 chip installed.
- You must specify at least one accelerator selection.
- You cannot specify the same accelerator name multiple times in the selections list.
- All accelerators in the selections must use the same runtime version.
count?
Type:
IResolvable | Accelerator
(optional)
The number of GPU accelerators specified for worker hosts in this fleet.
You must specify either
acceleratorCapabilities.count.maxorallowedInstanceTypeswhen using accelerator capabilities. If you don't specify a maximum count, AWS Deadline Cloud uses the instance types you specify inallowedInstanceTypesto determine the maximum number of accelerators.

.NET
Go
Java
Python
TypeScript