fix: accept a PipelineVariable for HyperparameterTuner random_seed

sagemaker-core/src/sagemaker/core/shapes/shapes.py

@@ -7438,7 +7438,7 @@ class HyperParameterTuningJobConfig(Base):
     parameter_ranges: Optional[ParameterRanges] = Unassigned()
     training_job_early_stopping_type: Optional[StrPipeVar] = Unassigned()
     tuning_job_completion_criteria: Optional[TuningJobCompletionCriteria] = Unassigned()
-    random_seed: Optional[int] = Unassigned()
+    random_seed: Optional[IntPipeVar] = Unassigned()
 
 
 class HyperParameterAlgorithmSpecification(Base):

sagemaker-core/src/sagemaker/core/tools/constants.py

@@ -147,4 +147,9 @@
     "ProcessingInstancePreference": {
         "InstanceCount": "IntPipeVar",
     },
+    # RandomSeed accepts a pipeline variable (e.g. a ParameterInteger) so tuning pipelines
+    # can parameterize reproducibility (issue #5614 / #6171).
+    "HyperParameterTuningJobConfig": {
+        "RandomSeed": "IntPipeVar",
+    },
 }

sagemaker-train/src/sagemaker/train/tuner.py

@@ -106,7 +106,7 @@ def __init__(
         completion_criteria_config: Optional[TuningJobCompletionCriteria] = None,
         early_stopping_type: Union[str, PipelineVariable] = "Off",
         model_trainer_name: Optional[str] = None,
-        random_seed: Optional[int] = None,
+        random_seed: Optional[Union[int, PipelineVariable]] = None,
         autotune: bool = False,
         hyperparameters_to_keep_static: Optional[List[str]] = None,
     ):
@@ -172,7 +172,8 @@ def __init__(
             model_trainer_name (str): A unique name to identify a model_trainer within the
                 hyperparameter tuning job, when more than one model_trainer is used with
                 the same tuning job (default: None).
-            random_seed (int): An initial value used to initialize a pseudo-random number generator.
+            random_seed (int or PipelineVariable): An initial value used to initialize a pseudo-random
+                number generator.
                 Setting a random seed will make the hyperparameter tuning search strategies to
                 produce more consistent configurations for the same tuning job.
             autotune (bool): Whether the parameter ranges or other unset settings of a tuning job
@@ -1081,7 +1082,8 @@ def create(
                 Can be either 'Auto' or 'Off' (default: 'Off'). If set to 'Off', early stopping
                 will not be attempted. If set to 'Auto', early stopping of some training jobs may
                 happen, but is not guaranteed to.
-            random_seed (int): An initial value used to initialize a pseudo-random number generator.
+            random_seed (int or PipelineVariable): An initial value used to initialize a pseudo-random
+                number generator.
                 Setting a random seed will make the hyperparameter tuning search strategies to
                 produce more consistent configurations for the same tuning job.
             autotune (bool): Whether the parameter ranges or other unset settings of a tuning job

sagemaker-train/tests/unit/train/test_tuner.py

@@ -261,6 +261,30 @@ def test_init_with_random_seed(self, mock_model_trainer, hyperparameter_ranges):
 
         assert tuner.random_seed == 42
 
+    def test_random_seed_accepts_pipeline_variable(self, mock_model_trainer, hyperparameter_ranges):
+        """Regression for #5614 / #6171.
+
+        ``random_seed`` must accept a pipeline variable (e.g. a ParameterInteger). Building the
+        tuning job config assigns it to ``HyperParameterTuningJobConfig.random_seed`` under
+        ``validate_assignment=True``; when that field was typed ``Optional[int]`` this raised
+        ``ValidationError: 1 validation error for HyperParameterTuningJobConfig``.
+        """
+        from sagemaker.core.workflow.parameters import ParameterInteger
+
+        seed = ParameterInteger(name="RandomState", default_value=42)
+        tuner = HyperparameterTuner(
+            model_trainer=mock_model_trainer,
+            objective_metric_name="accuracy",
+            hyperparameter_ranges=hyperparameter_ranges,
+            max_jobs=2,
+            max_parallel_jobs=1,
+            random_seed=seed,
+        )
+
+        assert tuner.random_seed is seed
+        config = tuner._build_tuning_job_config()
+        assert config.random_seed is seed
+
     def test_init_with_autotune(self, mock_model_trainer):
         """Test initialization with autotune enabled."""
         tuner = HyperparameterTuner(