Delen via

ComputeOperations Class

  • Reference

ComputeOperations.

This class should not be instantiated directly. Instead, use the compute attribute of an MLClient object.

Constructor

ComputeOperations(operation_scope: OperationScope, operation_config: OperationConfig, service_client: AzureMachineLearningWorkspaces, service_client_2024: AzureMachineLearningWorkspaces, **kwargs: Dict)

Parameters

Name Description
operation_scope
Required
<xref:azure.ai.ml._scope_dependent_operations.OperationScope>

Scope variables for the operations classes of an MLClient object.
operation_config
Required
<xref:azure.ai.ml._scope_dependent_operations.OperationConfig>

Common configuration for operations classes of an MLClient object.
service_client
Required
<xref:azure.ai.ml._restclient.v2023_02_01_preview.AzureMachineLearningWorkspaces>

Service client to allow end users to operate on Azure Machine Learning Workspace resources.
service_client_2024
Required

Methods

begin_attach

Attach a compute resource to the workspace.
begin_create_or_update

Create and register a compute resource.
begin_delete

Delete or detach a compute resource.
begin_restart

Restart a compute instance.
begin_start

Start a compute instance.
begin_stop

Stop a compute instance.
begin_update

Update a compute resource. Currently only valid for AmlCompute resource types.
enable_sso

Note

This is an experimental method, and may change at any time. Please see https://aka.ms/azuremlexperimental for more information.

enable sso for a compute instance.
get

Get a compute resource.
list

List computes of the workspace.
list_nodes

Retrieve a list of a compute resource's nodes.
list_sizes

List the supported VM sizes in a location.
list_usage

List the current usage information as well as AzureML resource limits for the given subscription and location.

begin_attach

Attach a compute resource to the workspace.

begin_attach(compute: Compute, **kwargs: Any) -> LROPoller[Compute]

Parameters

Name Description
compute
Required
Compute

The compute resource definition.

Returns

Type Description
LROPoller[Compute]

An instance of LROPoller that returns a Compute object once the long-running operation is complete.

Examples

Attaching a compute resource to the workspace.


   from azure.ai.ml.entities import AmlCompute

   compute_obj = AmlCompute(
       name=compute_name_2,
       tags={"key1": "value1", "key2": "value2"},
       min_instances=0,
       max_instances=10,
       idle_time_before_scale_down=100,
   )
   attached_compute = ml_client.compute.begin_attach(compute_obj)

begin_create_or_update

Create and register a compute resource.

begin_create_or_update(compute: Compute) -> LROPoller[Compute]

Parameters

Name Description
compute
Required
Compute

The compute resource definition.

Returns

Type Description
LROPoller[Compute]

An instance of LROPoller that returns a Compute object once the long-running operation is complete.

Examples

Creating and registering a compute resource.


   from azure.ai.ml.entities import AmlCompute

   compute_obj = AmlCompute(
       name=compute_name_1,
       tags={"key1": "value1", "key2": "value2"},
       min_instances=0,
       max_instances=10,
       idle_time_before_scale_down=100,
   )
   registered_compute = ml_client.compute.begin_create_or_update(compute_obj)

begin_delete

Delete or detach a compute resource.

begin_delete(name: str, *, action: str = 'Delete') -> LROPoller[None]

Parameters

Name Description
name
Required
str

The name of the compute resource.

Keyword-Only Parameters

Name Description
action

Action to perform. Possible values: ["Delete", "Detach"]. Defaults to "Delete".

Default value: Delete

Returns

Type Description
LROPoller[None]

A poller to track the operation status.

Examples

Delete compute example.


   ml_client.compute.begin_delete(compute_name_1, action="Detach")

   ml_client.compute.begin_delete(compute_name_2)

begin_restart

Restart a compute instance.

begin_restart(name: str) -> LROPoller[None]

Parameters

Name Description
name
Required
str

The name of the compute instance.

Returns

Type Description
LROPoller[None]

A poller to track the operation status.

Examples

Restarting a stopped compute instance.


   ml_client.compute.begin_restart(ci_name)

begin_start

Start a compute instance.

begin_start(name: str) -> LROPoller[None]

Parameters

Name Description
name
Required
str

The name of the compute instance.

Returns

Type Description
LROPoller[None]

A poller to track the operation status.

Examples

Starting a compute instance.


   ml_client.compute.begin_start(ci_name)

begin_stop

Stop a compute instance.

begin_stop(name: str) -> LROPoller[None]

Parameters

Name Description
name
Required
str

The name of the compute instance.

Returns

Type Description
LROPoller[None]

A poller to track the operation status.

Examples

Stopping a compute instance.


   ml_client.compute.begin_stop(ci_name)

begin_update

Update a compute resource. Currently only valid for AmlCompute resource types.

begin_update(compute: Compute) -> LROPoller[Compute]

Parameters

Name Description
compute
Required
Compute

The compute resource definition.

Returns

Type Description
LROPoller[Compute]

An instance of LROPoller that returns a Compute object once the long-running operation is complete.

Examples

Updating an AmlCompute resource.


   compute_obj = ml_client.compute.get("cpu-cluster")
   compute_obj.idle_time_before_scale_down = 200
   updated_compute = ml_client.compute.begin_update(compute_obj)

enable_sso

Note

This is an experimental method, and may change at any time. Please see https://aka.ms/azuremlexperimental for more information.

enable sso for a compute instance.

enable_sso(*, name: str, enable_sso: bool = True, **kwargs: Any) -> None

Keyword-Only Parameters

Name Description
name
str

Name of the compute instance.
enable_sso
bool

enable sso bool flag Default to True

Default value: True

get

Get a compute resource.

get(name: str) -> Compute

Parameters

Name Description
name
Required
str

Name of the compute resource.

Returns

Type Description
Compute

A Compute object.

Examples

Retrieving a compute resource from a workspace.


   cpu_cluster = ml_client.compute.get("cpu-cluster")

list

List computes of the workspace.

list(*, compute_type: str | None = None) -> Iterable[Compute]

Keyword-Only Parameters

Name Description
compute_type
Optional[str]

The type of the compute to be listed, case-insensitive. Defaults to AMLCompute.

Default value: None

Returns

Type Description
ItemPaged[Compute]

An iterator like instance of Compute objects.

Examples

Retrieving a list of the AzureML Kubernetes compute resources in a workspace.


   compute_list = ml_client.compute.list(compute_type="AMLK8s")  # cspell:disable-line

list_nodes

Retrieve a list of a compute resource's nodes.

list_nodes(name: str) -> Iterable[AmlComputeNodeInfo]

Parameters

Name Description
name
Required
str

Name of the compute resource.

Returns

Type Description
ItemPaged[AmlComputeNodeInfo]

An iterator-like instance of AmlComputeNodeInfo objects.

Examples

Retrieving a list of nodes from a compute resource.


   node_list = ml_client.compute.list_nodes(name="cpu-cluster")

list_sizes

List the supported VM sizes in a location.

list_sizes(*, location: str | None = None, compute_type: str | None = None) -> Iterable[VmSize]

Keyword-Only Parameters

Name Description
location
str

The location upon which virtual-machine-sizes is queried. Defaults to workspace location.

Default value: None
compute_type
Optional[str]

The type of the compute to be listed, case-insensitive. Defaults to AMLCompute.

Default value: None

Returns

Type Description
Iterable[VmSize]

An iterator over virtual machine size objects.

Examples

Listing the supported VM sizes in the workspace location.


   size_list = ml_client.compute.list_sizes()

list_usage

List the current usage information as well as AzureML resource limits for the given subscription and location.

list_usage(*, location: str | None = None) -> Iterable[Usage]

Keyword-Only Parameters

Name Description
location
Optional[str]

The location for which resource usage is queried. Defaults to workspace location.

Default value: None

Returns

Type Description
ItemPaged[Usage]

An iterator over current usage info objects.

Examples

Listing resource usage for the workspace location.


   usage_list = ml_client.compute.list_usage()