diff --git a/tests/modular_pipelines/stable_diffusion_xl/test_modular_pipeline_stable_diffusion_xl.py b/tests/modular_pipelines/stable_diffusion_xl/test_modular_pipeline_stable_diffusion_xl.py
index 1b9c791361..7f080bf8de 100644
--- a/tests/modular_pipelines/stable_diffusion_xl/test_modular_pipeline_stable_diffusion_xl.py
+++ b/tests/modular_pipelines/stable_diffusion_xl/test_modular_pipeline_stable_diffusion_xl.py
@@ -99,9 +99,9 @@ class SDXLModularTests:
 
         assert image.shape == expected_image_shape
 
-        assert np.abs(image_slice.flatten() - expected_slice).max() < expected_max_diff, (
-            "Image Slice does not match expected slice"
-        )
+        assert (
+            np.abs(image_slice.flatten() - expected_slice).max() < expected_max_diff
+        ), "Image Slice does not match expected slice"
 
 
 class SDXLModularIPAdapterTests:
@@ -114,20 +114,20 @@ class SDXLModularIPAdapterTests:
         parameters = blocks.input_names
 
         assert issubclass(self.pipeline_class, ModularIPAdapterMixin)
-        assert "ip_adapter_image" in parameters, (
-            "`ip_adapter_image` argument must be supported by the `__call__` method"
-        )
+        assert (
+            "ip_adapter_image" in parameters
+        ), "`ip_adapter_image` argument must be supported by the `__call__` method"
         assert "ip_adapter" in blocks.sub_blocks, "pipeline must contain an IPAdapter block"
 
         _ = blocks.sub_blocks.pop("ip_adapter")
         parameters = blocks.input_names
         intermediate_parameters = blocks.intermediate_input_names
-        assert "ip_adapter_image" not in parameters, (
-            "`ip_adapter_image` argument must be removed from the `__call__` method"
-        )
-        assert "ip_adapter_image_embeds" not in intermediate_parameters, (
-            "`ip_adapter_image_embeds` argument must be supported by the `__call__` method"
-        )
+        assert (
+            "ip_adapter_image" not in parameters
+        ), "`ip_adapter_image` argument must be removed from the `__call__` method"
+        assert (
+            "ip_adapter_image_embeds" not in intermediate_parameters
+        ), "`ip_adapter_image_embeds` argument must be supported by the `__call__` method"
 
     def _get_dummy_image_embeds(self, cross_attention_dim: int = 32):
         return torch.randn((1, 1, cross_attention_dim), device=torch_device)
@@ -203,9 +203,9 @@ class SDXLModularIPAdapterTests:
         max_diff_without_adapter_scale = np.abs(output_without_adapter_scale - output_without_adapter).max()
         max_diff_with_adapter_scale = np.abs(output_with_adapter_scale - output_without_adapter).max()
 
-        assert max_diff_without_adapter_scale < expected_max_diff, (
-            "Output without ip-adapter must be same as normal inference"
-        )
+        assert (
+            max_diff_without_adapter_scale < expected_max_diff
+        ), "Output without ip-adapter must be same as normal inference"
         assert max_diff_with_adapter_scale > 1e-2, "Output with ip-adapter must be different from normal inference"
 
         # 2. Multi IP-Adapter test cases
@@ -235,12 +235,12 @@ class SDXLModularIPAdapterTests:
             output_without_multi_adapter_scale - output_without_adapter
         ).max()
         max_diff_with_multi_adapter_scale = np.abs(output_with_multi_adapter_scale - output_without_adapter).max()
-        assert max_diff_without_multi_adapter_scale < expected_max_diff, (
-            "Output without multi-ip-adapter must be same as normal inference"
-        )
-        assert max_diff_with_multi_adapter_scale > 1e-2, (
-            "Output with multi-ip-adapter scale must be different from normal inference"
-        )
+        assert (
+            max_diff_without_multi_adapter_scale < expected_max_diff
+        ), "Output without multi-ip-adapter must be same as normal inference"
+        assert (
+            max_diff_with_multi_adapter_scale > 1e-2
+        ), "Output with multi-ip-adapter scale must be different from normal inference"
 
 
 class SDXLModularControlNetTests:
@@ -253,9 +253,9 @@ class SDXLModularControlNetTests:
         parameters = blocks.input_names
 
         assert "control_image" in parameters, "`control_image` argument must be supported by the `__call__` method"
-        assert "controlnet_conditioning_scale" in parameters, (
-            "`controlnet_conditioning_scale` argument must be supported by the `__call__` method"
-        )
+        assert (
+            "controlnet_conditioning_scale" in parameters
+        ), "`controlnet_conditioning_scale` argument must be supported by the `__call__` method"
 
     def _modify_inputs_for_controlnet_test(self, inputs: Dict[str, Any]):
         controlnet_embedder_scale_factor = 2
@@ -301,9 +301,9 @@ class SDXLModularControlNetTests:
         max_diff_without_controlnet_scale = np.abs(output_without_controlnet_scale - output_without_controlnet).max()
         max_diff_with_controlnet_scale = np.abs(output_with_controlnet_scale - output_without_controlnet).max()
 
-        assert max_diff_without_controlnet_scale < expected_max_diff, (
-            "Output without controlnet must be same as normal inference"
-        )
+        assert (
+            max_diff_without_controlnet_scale < expected_max_diff
+        ), "Output without controlnet must be same as normal inference"
         assert max_diff_with_controlnet_scale > 1e-2, "Output with controlnet must be different from normal inference"
 
     def test_controlnet_cfg(self):
@@ -383,26 +383,6 @@ class SDXLModularPipelineFastTests(
     def test_inference_batch_single_identical(self):
         super().test_inference_batch_single_identical(expected_max_diff=3e-3)
 
-    @require_torch_accelerator
-    def test_stable_diffusion_xl_offloads(self):
-        pipes = []
-        sd_pipe = self.get_pipeline().to(torch_device)
-        pipes.append(sd_pipe)
-
-        cm = ComponentsManager()
-        cm.enable_auto_cpu_offload(device=torch_device)
-        sd_pipe = self.get_pipeline(components_manager=cm)
-        pipes.append(sd_pipe)
-
-        image_slices = []
-        for pipe in pipes:
-            inputs = self.get_dummy_inputs(torch_device)
-            image = pipe(**inputs, output="images")
-
-            image_slices.append(image[0, -3:, -3:, -1].flatten())
-
-        assert np.abs(image_slices[0] - image_slices[1]).max() < 1e-3
-
     def test_stable_diffusion_xl_save_from_pretrained(self):
         pipes = []
         sd_pipe = self.get_pipeline().to(torch_device)
diff --git a/tests/modular_pipelines/test_modular_pipelines_common.py b/tests/modular_pipelines/test_modular_pipelines_common.py
index 5a8227f5ce..684ba3c583 100644
--- a/tests/modular_pipelines/test_modular_pipelines_common.py
+++ b/tests/modular_pipelines/test_modular_pipelines_common.py
@@ -320,11 +320,40 @@ class ModularPipelineTesterMixin:
                 assert images.shape[0] == batch_size * num_images_per_prompt
 
     @require_accelerator
-    def test_components_auto_cpu_offload(self):
+    def test_components_auto_cpu_offload_inference_consistent(self):
         base_pipe = self.get_pipeline().to(torch_device)
-        for component in base_pipe.components:
-            assert component.device == torch_device
 
         cm = ComponentsManager()
         cm.enable_auto_cpu_offload(device=torch_device)
         offload_pipe = self.get_pipeline(components_manager=cm)
+
+        image_slices = []
+        for pipe in [base_pipe, offload_pipe]:
+            inputs = self.get_dummy_inputs(torch_device)
+            image = pipe(**inputs, output="images")
+
+            image_slices.append(image[0, -3:, -3:, -1].flatten())
+
+        assert np.abs(image_slices[0] - image_slices[1]).max() < 1e-3
+
+    def test_save_from_pretrained(self):
+        pipes = []
+        base_pipe = self.get_pipeline().to(torch_device)
+        pipes.append(base_pipe)
+
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            base_pipe.save_pretrained(tmpdirname)
+            pipe = ModularPipeline.from_pretrained(tmpdirname).to(torch_device)
+            pipe.load_default_components(torch_dtype=torch.float32)
+            pipe.to(torch_device)
+
+        pipes.append(pipe)
+
+        image_slices = []
+        for pipe in pipes:
+            inputs = self.get_dummy_inputs(torch_device)
+            image = pipe(**inputs, output="images")
+
+            image_slices.append(image[0, -3:, -3:, -1].flatten())
+
+        assert np.abs(image_slices[0] - image_slices[1]).max() < 1e-3