Workload Class

A WorkloadClass resource specifies the hardware resources assigned to a Job that will be executed. Specifically, the resource specifies memory, CPU, and GPU allocations. Workload classes are not static and can be customized during the creation of objects that use the WorkloadClass resource. All workload classes can be viewed by navigating to the Workload Classes section of the modela-catalog namespace.

Relationships to Other Resources

All DataProduct resources have a default WorkloadClass.
All Dataset resources are assigned a WorkloadClass for data processing.
All Model resources are assigned a WorkloadClass during training.

WorkloadClass API Reference

WorkloadClass define a template for a job.

Name	Type	Description	Required
apiVersion	string	catalog.modela.ai/v1alpha1	true
kind	string	WorkloadClass	true
metadata	object	Refer to the Kubernetes API documentation for the fields of the `metadata` field.	true
spec	object	WorkloadClassSpec defines the specification of a workload class.	true

WorkloadClass.spec

^{↩ Parent}

WorkloadClassSpec defines the specification of a workload class.

Name	Type	Description	Required
resourcesTemplate	object	ResourceSpec specifies the amount of resources that will be allocated to a workload	false

WorkloadClass.spec.resourcesTemplate

^{↩ Parent}

ResourceSpec specifies the amount of resources that will be allocated to a workload

Name	Type	Description	Required
cpuImage	object	Reference to the managed CPU trainer image, used internally	false
gpuImage	object	Reference to the managed GPU trainer image, used internally	false
requirements	object	The custom resource requirements for the workload, which are used if `WorkloadName` is not set	false
workloadName	string	If this resource is based on the workload, this field contain the name of the workload. The name of a WorkloadClass. The system will use the resource requirements described by the WorkloadClass	false

WorkloadClass.spec.resourcesTemplate.cpuImage

^{↩ Parent}

Reference to the managed CPU trainer image, used internally

Name	Type	Description	Required
apiVersion	string	API version of the referent.	false
fieldPath	string	If referring to a piece of an object instead of an entire object, this string should contain a valid JSON/Go field access statement, such as desiredState.manifest.containers[2]. For example, if the object reference is to a container within a pod, this would take on a value like: "spec.containers{name}" (where "name" refers to the name of the container that triggered the event) or if no container name is specified "spec.containers[2]" (container with index 2 in this pod). This syntax is chosen only to have some well-defined way of referencing a part of an object. TODO: this design is not final and this field is subject to change in the future.	false
kind	string	Kind of the referent. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds	false
name	string	Name of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#names	false
namespace	string	Namespace of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/namespaces/	false
resourceVersion	string	Specific resourceVersion to which this reference is made, if any. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#concurrency-control-and-consistency	false
uid	string	UID of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#uids	false

WorkloadClass.spec.resourcesTemplate.gpuImage

^{↩ Parent}

Reference to the managed GPU trainer image, used internally

Name	Type	Description	Required
apiVersion	string	API version of the referent.	false
fieldPath	string	If referring to a piece of an object instead of an entire object, this string should contain a valid JSON/Go field access statement, such as desiredState.manifest.containers[2]. For example, if the object reference is to a container within a pod, this would take on a value like: "spec.containers{name}" (where "name" refers to the name of the container that triggered the event) or if no container name is specified "spec.containers[2]" (container with index 2 in this pod). This syntax is chosen only to have some well-defined way of referencing a part of an object. TODO: this design is not final and this field is subject to change in the future.	false
kind	string	Kind of the referent. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds	false
name	string	Name of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#names	false
namespace	string	Namespace of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/namespaces/	false
resourceVersion	string	Specific resourceVersion to which this reference is made, if any. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#concurrency-control-and-consistency	false
uid	string	UID of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#uids	false

WorkloadClass.spec.resourcesTemplate.requirements

^{↩ Parent}

The custom resource requirements for the workload, which are used if WorkloadName is not set

Name	Type	Description	Required
limits	map[string]int or string	Limits describes the maximum amount of compute resources allowed. More info: https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/	false
requests	map[string]int or string	Requests describes the minimum amount of compute resources required. If Requests is omitted for a container, it defaults to Limits if that is explicitly specified, otherwise to an implementation-defined value. More info: https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/	false