fix: [Bug] Pipeline parameters (ParameterInteger, ParameterString) fail in ModelTrain (5504)

[aviruthen]

Description

Add tests for PipelineVariable support in ModelTrainer hyperparameters

Problem

The safe_serialize function in sagemaker-train/src/sagemaker/train/utils.py was previously unable to handle PipelineVariable objects (e.g., ParameterInteger, ParameterString) passed as hyperparameter values, causing a TypeError: Object of type ParameterInteger is not JSON serializable when building pipelines.

While this bug has already been fixed in the codebase (the safe_serialize function now has an explicit elif isinstance(data, PipelineVariable): return data branch), there was a test gap: the existing test_model_trainer_pipeline_variable.py tested PipelineVariable for training_image, algorithm_name, training_input_mode, and environment fields but did NOT test PipelineVariable in hyperparameters — which is the exact scenario from the bug report.

Changes

Added the following tests to test_model_trainer_pipeline_variable.py:

In TestModelTrainerPipelineVariableAcceptance:

• test_hyperparameters_accept_parameter_integer — verifies ParameterInteger can be used as a hyperparameter value
• test_hyperparameters_accept_parameter_string — verifies ParameterString can be used as a hyperparameter value
• test_hyperparameters_accept_mixed_pipeline_and_static_values — verifies a mix of PipelineVariable and static values works

New TestSafeSerializePipelineVariable class:

• test_safe_serialize_with_parameter_integer — verifies safe_serialize returns ParameterInteger as-is
• test_safe_serialize_with_parameter_string — verifies safe_serialize returns ParameterString as-is
• test_safe_serialize_with_parameter_float — verifies safe_serialize returns ParameterFloat as-is
• test_safe_serialize_with_plain_string — regression test for plain string passthrough
• test_safe_serialize_with_int — regression test for integer JSON serialization
• test_safe_serialize_with_dict — regression test for dict JSON serialization

Testing

All new tests validate the fix is properly covered and will prevent regressions.

Related Issue

Related issue: 5504

Changes Made

The bug described in the issue has already been fixed in the current codebase. The safe_serialize function in sagemaker-train/src/sagemaker/train/utils.py already handles PipelineVariable objects with an explicit elif isinstance(data, PipelineVariable): return data branch (line ~176). The serialize function in sagemaker-core/src/sagemaker/core/utils/utils.py also handles PipelineVariable by returning it as-is. Both functions correctly prevent the TypeError: Object of type ParameterInteger is not JSON serializable error. However, there is a test gap: the existing test_model_trainer_pipeline_variable.py tests PipelineVariable for training_image, algorithm_name, training_input_mode, and environment fields but does NOT test PipelineVariable in hyperparameters — which is the exact scenario from the bug report. We should add tests to ensure this scenario stays covered.

AI-Generated PR

This PR was automatically generated by the PySDK Issue Agent.

• Confidence score: 92%
• Classification: bug
• SDK version target: V3

Merge Checklist

• Changes are backward compatible
• Commit message follows prefix: description format
• Unit tests added/updated
• Integration tests added (if applicable)
• Documentation updated (if applicable)