parallel Package

Classes

ParallelJob

Parallel job.

RunFunction

Run Function.

Functions

parallel_run_function

Create a Parallel object which can be used inside dsl.pipeline as a function and can also be created as a standalone parallel job.

For an example of using ParallelRunStep, see the notebook https://aka.ms/parallel-example-notebook

Note

To use parallel_run_function:

Create a <xref:azure.ai.ml.entities._builders.Parallel> object to specify how parallel run is performed,

with parameters to control batch size,number of nodes per compute target, and a

reference to your custom Python script.

Build pipeline with the parallel object as a function. defines inputs and

outputs for the step.

Sumbit the pipeline to run.


   from azure.ai.ml import Input, Output, parallel

   parallel_run = parallel_run_function(
       name="batch_score_with_tabular_input",
       display_name="Batch Score with Tabular Dataset",
       description="parallel component for batch score",
       inputs=dict(
           job_data_path=Input(
               type=AssetTypes.MLTABLE,
               description="The data to be split and scored in parallel",
           ),
           score_model=Input(
               type=AssetTypes.URI_FOLDER, description="The model for batch score."
           ),
       ),
       outputs=dict(job_output_path=Output(type=AssetTypes.MLTABLE)),
       input_data="${{inputs.job_data_path}}",
       max_concurrency_per_instance=2,  # Optional, default is 1
       mini_batch_size="100",  # optional
       mini_batch_error_threshold=5,  # Optional, allowed failed count on mini batch items, default is -1
       logging_level="DEBUG",  # Optional, default is INFO
       error_threshold=5,  # Optional, allowed failed count totally, default is -1
       retry_settings=dict(max_retries=2, timeout=60),  # Optional
       task=RunFunction(
           code="./src",
           entry_script="tabular_batch_inference.py",
           environment=Environment(
               image="mcr.microsoft.com/azureml/openmpi3.1.2-ubuntu18.04",
               conda_file="./src/environment_parallel.yml",
           ),
           program_arguments="--model ${{inputs.score_model}}",
           append_row_to="${{outputs.job_output_path}}",  # Optional, if not set, summary_only
       ),
   )
parallel_run_function(*, name: str | None = None, description: str | None = None, tags: Dict | None = None, properties: Dict | None = None, display_name: str | None = None, experiment_name: str | None = None, compute: str | None = None, retry_settings: BatchRetrySettings | None = None, environment_variables: Dict | None = None, logging_level: str | None = None, max_concurrency_per_instance: int | None = None, error_threshold: int | None = None, mini_batch_error_threshold: int | None = None, task: RunFunction | None = None, mini_batch_size: str | None = None, partition_keys: List | None = None, input_data: str | None = None, inputs: Dict | None = None, outputs: Dict | None = None, instance_count: int | None = None, instance_type: str | None = None, docker_args: str | None = None, shm_size: str | None = None, identity: ManagedIdentityConfiguration | AmlTokenConfiguration | UserIdentityConfiguration | None = None, is_deterministic: bool = True, **kwargs: Any) -> Parallel

Keyword-Only Parameters

Name Description
name
str

Name of the parallel job or component created.

description
str

A friendly description of the parallel.

tags

Tags to be attached to this parallel.

properties

The asset property dictionary.

display_name
str

A friendly name.

experiment_name
str

Name of the experiment the job will be created under, if None is provided, default will be set to current directory name. Will be ignored as a pipeline step.

compute
str

The name of the compute where the parallel job is executed (will not be used if the parallel is used as a component/function).

retry_settings

Parallel component run failed retry

environment_variables

A dictionary of environment variables names and values. These environment variables are set on the process where user script is being executed.

logging_level
str

A string of the logging level name, which is defined in 'logging'. Possible values are 'WARNING', 'INFO', and 'DEBUG'. (optional, default value is 'INFO'.) This value could be set through PipelineParameter.

max_concurrency_per_instance
int

The max parallellism that each compute instance has.

error_threshold
int

The number of record failures for Tabular Dataset and file failures for File Dataset that should be ignored during processing. If the error count goes above this value, then the job will be aborted. Error threshold is for the entire input rather than the individual mini-batch sent to run() method. The range is [-1, int.max]. -1 indicates ignore all failures during processing

mini_batch_error_threshold
int

The number of mini batch processing failures should be ignored

task

The parallel task

mini_batch_size
str

For FileDataset input, this field is the number of files a user script can process in one run() call. For TabularDataset input, this field is the approximate size of data the user script can process in one run() call. Example values are 1024, 1024KB, 10MB, and 1GB. (optional, default value is 10 files for FileDataset and 1MB for TabularDataset.) This value could be set through PipelineParameter.

partition_keys

The keys used to partition dataset into mini-batches. If specified, the data with the same key will be partitioned into the same mini-batch. If both partition_keys and mini_batch_size are specified, the partition keys will take effect. The input(s) must be partitioned dataset(s), and the partition_keys must be a subset of the keys of every input dataset for this to work

input_data
str

The input data.

inputs

A dict of inputs used by this parallel.

outputs

The outputs of this parallel

instance_count
int

Optional number of instances or nodes used by the compute target. Defaults to 1

instance_type
str

Optional type of VM used as supported by the compute target..

docker_args
str

Extra arguments to pass to the Docker run command. This would override any parameters that have already been set by the system, or in this section. This parameter is only supported for Azure ML compute types.

shm_size
str

Size of the docker container's shared memory block. This should be in the format of (number)(unit) where number as to be greater than 0 and the unit can be one of b(bytes), k(kilobytes), m(megabytes), or g(gigabytes).

identity

Identity that PRS job will use while running on compute.

is_deterministic

Specify whether the parallel will return same output given same input. If a parallel (component) is deterministic, when use it as a node/step in a pipeline, it will reuse results from a previous submitted job in current workspace which has same inputs and settings. In this case, this step will not use any compute resource. Defaults to True, specify is_deterministic=False if you would like to avoid such reuse behavior, defaults to True.

Default value: True

Returns

Type Description

The parallel node