[ML] Fix ml_client.jobs.download for BatchJob (Azure#29316)

* fix: Avoid matching BatchJob child by name during Job.downlaod

    When downloading a BatchJob, the SDK needed to do a manual search
    through the BatchJob's children to find the output.

    The SDK attempted to make the child by name, but a recent service
    change caused that search to fail.

    BatchJobs only have one child, so matching by name is largely
    unnecessary anyway.

* Update CHANGELOG.md

Author: kdestin

sdk/ml/azure-ai-ml/CHANGELOG.md

@@ -29,6 +29,7 @@
 - Error message improvement when a local path fails to match with data asset type.
 - Error message improvement when an asset does not exist in a registry
 - Fix an issue when submit spark pipeline job with referring a registered component
+- Fix an issue that prevented Job.download from downloading the output of a BatchJob
 
 ### Other Changes
 

sdk/ml/azure-ai-ml/azure/ai/ml/constants/_common.py

@@ -115,7 +115,6 @@
 REF_DOC_YAML_SCHEMA_ERROR_MSG_FORMAT = "\nVisit this link to refer to the {} schema if needed: {}."
 STORAGE_AUTH_MISMATCH_ERROR = "AuthorizationPermissionMismatch"
 SWEEP_JOB_BEST_CHILD_RUN_ID_PROPERTY_NAME = "best_child_run_id"
-BATCH_JOB_CHILD_RUN_NAME = "batchscoring"
 BATCH_JOB_CHILD_RUN_OUTPUT_NAME = "score"
 DEFAULT_ARTIFACT_STORE_OUTPUT_NAME = "default"
 DEFAULT_EXPERIMENT_NAME = "Default"

sdk/ml/azure-ai-ml/azure/ai/ml/operations/_job_operations.py

@@ -50,7 +50,6 @@
 from azure.ai.ml.constants._common import (
     API_URL_KEY,
     AZUREML_RESOURCE_PROVIDER,
-    BATCH_JOB_CHILD_RUN_NAME,
     BATCH_JOB_CHILD_RUN_OUTPUT_NAME,
     COMMON_RUNTIME_ENV_VAR,
     DEFAULT_ARTIFACT_STORE_OUTPUT_NAME,
@@ -831,12 +830,13 @@ def _get_named_output_uri(
     def _get_batch_job_scoring_output_uri(self, job_name: str) -> Optional[str]:
         uri = None
         # Download scoring output, which is the "score" output of the child job named "batchscoring"
+        # Batch Jobs are pipeline jobs with only one child, so this should terminate after an iteration
         for child in self._runs_operations.get_run_children(job_name):
-            if child.properties.get("azureml.moduleName", None) == BATCH_JOB_CHILD_RUN_NAME:
-                uri = self._get_named_output_uri(child.name, BATCH_JOB_CHILD_RUN_OUTPUT_NAME).get(
-                    BATCH_JOB_CHILD_RUN_OUTPUT_NAME, None
-                )
-                # After the correct child is found, break to prevent unnecessary looping
+            uri = self._get_named_output_uri(child.name, BATCH_JOB_CHILD_RUN_OUTPUT_NAME).get(
+                BATCH_JOB_CHILD_RUN_OUTPUT_NAME, None
+            )
+            # After the correct child is found, break to prevent unnecessary looping
+            if uri is not None:
                 break
         return uri