Batch component added for deployment and inference

assets/common/components/batch_deploy_model/asset.yaml

@@ -0,0 +1,3 @@
+type: component
+spec: spec.yaml
+categories: ["Models"]

assets/common/components/batch_deploy_model/spec.yaml

@@ -0,0 +1,191 @@
+$schema: https://azuremlschemas.azureedge.net/latest/commandComponent.schema.json
+name: batch_deploy_model
+version: 0.0.1
+type: command
+
+is_deterministic: True
+
+display_name: Batch deploy model
+description:
+ Batch deploy a model to a workspace. The component works on compute with [MSI](https://learn.microsoft.com/en-us/azure/machine-learning/how-to-create-manage-compute-instance?tabs=python) attached.
+
+environment: azureml://registries/azureml/environments/python-sdk-v2/versions/6
+
+code: ../../src
+command: >-
+ python batch_deploy.py
+ $[[--registration_details ${{inputs.registration_details}}]]
+ $[[--model_id ${{inputs.model_id}}]]
+ $[[--inference_payload ${{inputs.inference_payload}}]]
+ $[[--endpoint_name ${{inputs.endpoint_name}}]]
+ $[[--deployment_name ${{inputs.deployment_name}}]]
+ $[[--compute_name ${{inputs.compute_name}}]]
+ $[[--size ${{inputs.size}}]]
+ $[[--min_instances ${{inputs.min_instances}}]]
+ $[[--max_instances ${{inputs.max_instances}}]]
+ $[[--idle_time_before_scale_down ${{inputs.idle_time_before_scale_down}}]]
+ $[[--output_file_name ${{inputs.output_file_name}}]]
+ $[[--max_concurrency_per_instance ${{inputs.max_concurrency_per_instance}}]]
+ $[[--error_threshold ${{inputs.error_threshold}}]]
+ $[[--max_retries ${{inputs.max_retries}}]]
+ $[[--timeout ${{inputs.timeout}}]]
+ $[[--logging_level ${{inputs.logging_level}}]]
+ $[[--mini_batch_size ${{inputs.mini_batch_size}}]]
+ $[[--instance_count ${{inputs.instance_count}}]]
+ --model_deployment_details ${{outputs.model_deployment_details}}
+
+inputs:
+ # Output of registering component
+ registration_details:
+ type: uri_file
+ optional: true
+ description: Json file that contains the ID of registered model to be deployed
+
+ model_id:
+ type: string
+ optional: true 
+ description: |
+ Asset ID of the model registered in workspace/registry.
+ Registry - azureml://registries/<registry-name>/models/<model-name>/versions/<version>
+ Workspace - azureml:<model-name>:<version>
+
+ inference_payload:
+ type: uri_folder
+ optional: true
+ description: Folder containing files used to validate deployment
+
+ endpoint_name:
+ type: string
+ optional: true
+ description: Name of the endpoint
+
+ deployment_name:
+ type: string
+ optional: true
+ default: default
+ description: Name of the deployment
+
+ compute_name:
+ type: string
+ optional: true
+ default: cpu-cluster
+ description: Name of the deployment
+
+ size:
+ type: string
+ optional: true
+ enum:
+ - Standard_DS1_v2
+ - Standard_DS2_v2
+ - Standard_DS3_v2
+ - Standard_DS4_v2
+ - Standard_DS5_v2
+ - Standard_F2s_v2
+ - Standard_F4s_v2
+ - Standard_F8s_v2
+ - Standard_F16s_v2
+ - Standard_F32s_v2
+ - Standard_F48s_v2
+ - Standard_F64s_v2
+ - Standard_F72s_v2
+ - Standard_FX24mds
+ - Standard_FX36mds
+ - Standard_FX48mds
+ - Standard_E2s_v3
+ - Standard_E4s_v3
+ - Standard_E8s_v3
+ - Standard_E16s_v3
+ - Standard_E32s_v3
+ - Standard_E48s_v3
+ - Standard_E64s_v3
+ - Standard_NC4as_T4_v3
+ - Standard_NC6s_v2
+ - Standard_NC6s_v3
+ - Standard_NC8as_T4_v3
+ - Standard_NC12s_v2
+ - Standard_NC12s_v3
+ - Standard_NC16as_T4_v3
+ - Standard_NC24s_v2
+ - Standard_NC24s_v3
+ - Standard_NC24rs_v3
+ - Standard_NC64as_T4_v3
+ - Standard_ND40rs_v2
+ - Standard_ND96asr_v4
+ - Standard_ND96amsr_A100_v4
+ default: Standard_NC24s_v3
+ description: Compute instance size to deploy model. Make sure that instance type is available and have enough quota available.
+
+ min_instances:
+ type: integer
+ optional: true
+ default: 0
+ description: Minimum instances of the compute cluster
+
+ max_instances:
+ type: integer
+ optional: true
+ default: 3
+ description: Maximum instances of the compute cluster
+
+ idle_time_before_scale_down:
+ type: integer
+ optional: true
+ default: 120
+ description: Idle time before scale down of the compute cluster
+
+ output_file_name:
+ type: string
+ optional: true
+ default: predictions.csv
+ description: Name of the output file
+
+ max_concurrency_per_instance:
+ type: integer
+ optional: true
+ default: 1
+ description: Maximum number of parallelism per instance
+
+ error_threshold:
+ type: integer
+ optional: true
+ default: -1
+ description: Error threshold before batch inference is aborted
+
+ max_retries:
+ type: integer
+ optional: true
+ default: 3
+ description: Number of retries in failure
+
+ timeout:
+ type: integer
+ optional: true
+ default: 30
+ description: Timeout in seconds
+
+ logging_level:
+ type: string
+ optional: true
+ default: info
+ description: Logging level for batch inference operation
+
+ mini_batch_size:
+ type: integer
+ optional: true
+ default: 10
+ description: Size of the mini-batch passed to each batch invocation
+
+ instance_count:
+ type: integer
+ optional: true
+ default: 1
+ description: Number of compute instances to deploy model
+
+outputs:
+ model_deployment_details:
+ type: uri_file
+ description: Json file to which deployment details will be written
+
+tags:
+ Preview: ""
+ Internal: ""