More TF fixes (#28081)

* More build_in_name_scope()

* Make sure we set the save spec now we don't do it with dummies anymore

* make fixup

tests/models/gpt2/test_modeling_tf_gpt2.py

@@ -445,7 +445,7 @@ class TFGPT2ModelTest(TFModelTesterMixin, TFCoreModelTesterMixin, PipelineTester
                 continue
 
             model = model_class(config)
-            model.build()
+            model.build_in_name_scope()
 
             onnx_model_proto, _ = tf2onnx.convert.from_keras(model, opset=self.onnx_min_opset)