Extend `PipelineVariable` support in `ModelTrainer` to match `Estimator` (V2 -> V3 migration blocker)

[admivsn]

PySDK Version

• PySDK V2 (2.x)
• PySDK V3 (3.x)

Describe the bug
In SageMaker V2 we had Estimator which had vast support for pipeline variables. For example image_uri:

class Estimator(EstimatorBase):
 """A generic Estimator to train using any supplied algorithm.

 This class is designed for use with algorithms that don't have their own, custom class.
 """

 def __init__(
 self,
 image_uri: Union[str, PipelineVariable],

In SageMaker V3, we replace Estimator with ModelTrainer, however the support for pipeline variables is lacking. For example:

class ModelTrainer(BaseModel):
 """Class that trains a model using AWS SageMaker.

 ...
 """
 ...
 training_image: Optional[str] = None

Please help us to migrate to SageMaker V3 by adding equivalent support for pipeline variables to ModelTrainer, or let us know when we could expect the support to be added.

Alternatively, happy to work on this given some guidance.

To reproduce

from sagemaker.core.workflow.parameters import ParameterString
from sagemaker.core.workflow.pipeline_context import PipelineSession
from sagemaker.train import ModelTrainer

pipeline_session = PipelineSession(default_bucket="my-bucket")
training_image = ParameterString(name="training_image")

trainer = ModelTrainer(
 training_image=training_image,
 sagemaker_session=pipeline_session
)

ValidationError: 1 validation error for ModelTrainer
training_image
 Input should be a valid string [type=string_type, input_value=ParameterString(name='tra...g'>, default_value=None), input_type=ParameterString]

Expected behavior
ModelTrainer should accept PipelineVariable for training_image (and other relevant fields), matching V2 Estimator behaviour.