adding get and set lr for optimizer (#13661)

### Description
Exposing get and set Learning rate for optimizer


### Motivation and Context
you can now set learning rate for optimizer.

orttraining/orttraining/python/orttraining_pybind_state.cc

@@ -897,6 +897,12 @@ void addObjectMethodsForTraining(py::module& m, ExecutionProviderRegistrationFn
                           model->NamedParameters(), session_option,
                           GetTrainingORTEnv(), std::vector<std::shared_ptr<IExecutionProvider>>());
                     }))
+      .def("set_learning_rate", [](onnxruntime::training::api::Optimizer* optimizer, float lr) -> void {
+        ORT_THROW_IF_ERROR(optimizer->SetLearningRate(lr));
+      })
+      .def("get_learning_rate", [](onnxruntime::training::api::Optimizer* optimizer) -> float {
+        return optimizer->GetLearningRate();
+      })
       .def("optimizer_step", [](onnxruntime::training::api::Optimizer* optimizer) -> void {
         ORT_THROW_IF_ERROR(optimizer->Step());
       });

orttraining/orttraining/python/training/api/optimizer.py

@@ -22,3 +22,15 @@ class Optimizer:
         Run Optimizer Step.
         """
         self._optimizer.optimizer_step()
+
+    def set_learning_rate(self, learning_rate: float) -> None:
+        """
+        Set Learning Rate.
+        """
+        self._optimizer.set_learning_rate(learning_rate)
+
+    def get_learning_rate(self) -> float:
+        """
+        Get Learning Rate.
+        """
+        return self._optimizer.get_learning_rate()

orttraining/orttraining/test/python/orttraining_test_python_bindings.py

@@ -142,6 +142,32 @@ def test_optimizer_step():
         # TODO : Check if parameters changed from before and after optimizer step.
 
 
+def test_get_and_set_lr():
+    # Initialize Models
+    simple_model, onnx_model, optimizer_model, _, _ = _create_training_models()
+
+    with tempfile.TemporaryDirectory() as temp_dir:
+        # Save models & checkpoint files to load them later.
+        checkpoint_file_path, model_file_path, optimizer_file_path = _get_test_models_path(
+            temp_dir, simple_model, onnx_model, optimizer_model=optimizer_model
+        )
+        # Create Checkpoint State.
+        state = CheckpointState(checkpoint_file_path)
+        # Create a Module and Optimizer.
+        model = Module(model_file_path, state)
+        optimizer = Optimizer(optimizer_file_path, model)
+
+        # Test get and set learning rate.
+        lr = optimizer.get_learning_rate()
+        assert round(lr, 3) == 0.001
+
+        optimizer.set_learning_rate(0.5)
+        new_lr = optimizer.get_learning_rate()
+
+        assert np.isclose(new_lr, 0.5)
+        assert lr != new_lr
+
+
 def test_training_module_checkpoint():
     # Initialize Models
     simple_model, onnx_model, _, _, _ = _create_training_models()