[PT FE]: fix segfault when resolving nested dict as model input (openvinotoolkit#24719)

### Details:
 - *item1*
 - *...*

### Tickets:
 - *ticket-id*

Author: eaidova

src/frontends/pytorch/src/transforms/dict_resolver.cpp

@@ -21,7 +21,7 @@ using namespace ov::op;
 
 bool DictParameterResolver::run_on_model(const std::shared_ptr<Model>& model) {
     bool changed = false;
-    const auto& parameters = model->get_parameters();
+    const auto parameters = model->get_parameters();
     ParameterVector new_params;
 
     for (const auto& p : parameters) {

tests/layer_tests/ovc_python_api_tests/test_pytorch.py

@@ -1016,6 +1016,28 @@ def forward(self, x):
         "compress_to_fp16": False}
 
 
+def create_pytorch_module_with_nested_dict_input(tmp_dir):
+    class PTModel(torch.nn.Module):
+        def forward(self, a, b):
+            return a["1"] * a["2"] + b
+
+    net = PTModel()
+    a1 = ov.opset10.parameter(PartialShape([-1]), dtype=np.float32)
+    a2 = ov.opset10.parameter(PartialShape([-1]), dtype=np.float32)
+    b = ov.opset10.parameter(PartialShape([-1]), dtype=np.float32)
+    mul = ov.opset10.multiply(a1, a2)
+    add = ov.opset10.add(mul, b)
+    ref_model = Model([add], [a1, a2, b], "test")
+    return net, ref_model, {
+        "example_input": (
+            {
+            "1": torch.tensor([1, 2], dtype=torch.float32), 
+            "2": torch.tensor([3, 4], dtype=torch.float32)
+            }, 
+            torch.tensor([5, 6], dtype=torch.float32)
+        )}
+
+
 class TestMoConvertPyTorch(CommonMOConvertTest):
     test_data = [
         create_pytorch_nn_module_case1,
@@ -1067,7 +1089,8 @@ class TestMoConvertPyTorch(CommonMOConvertTest):
         create_pytorch_module_with_nested_inputs5,
         create_pytorch_module_with_nested_inputs6,
         create_pytorch_module_with_nested_list_and_single_input,
-        create_pytorch_module_with_single_input_as_list
+        create_pytorch_module_with_single_input_as_list,
+        create_pytorch_module_with_nested_dict_input
     ]
 
     @pytest.mark.parametrize("create_model", test_data)