[test_models_transformer_hunyuan_video] help us test torch.compile() for impactful models

tests/models/transformers/test_models_transformer_hunyuan_video.py

@@ -18,6 +18,7 @@
 
 from diffusers import HunyuanVideoTransformer3DModel
 from diffusers.utils.testing_utils import enable_full_determinism, torch_device
+from diffusers.utils.testing_utils import require_torch_gpu, require_torch_2, is_torch_compile, slow
 
 from ..test_modeling_common import ModelTesterMixin
 
@@ -88,6 +89,25 @@ def prepare_init_args_and_inputs_for_common(self):
     def test_gradient_checkpointing_is_applied(self):
         expected_set = {"HunyuanVideoTransformer3DModel"}
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
+
+    def test_gradient_checkpointing_is_applied(self):
+        expected_set = {"HunyuanVideoTransformer3DModel"}
+        super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
+
+    @require_torch_gpu
+    @require_torch_2
+    @is_torch_compile
+    @slow    
+    def test_torch_compile_recompilation_and_graph_break(self):
+        torch._dynamo.reset()
+        init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()
+
+        model = self.model_class(**init_dict).to(torch_device)
+        model = torch.compile(model, fullgraph=True)
+
+        with torch._dynamo.config.patch(error_on_recompile=True), torch.no_grad():
+            _ = model(**inputs_dict)
+            _ = model(**inputs_dict)
 
 
 class HunyuanSkyreelsImageToVideoTransformer3DTests(ModelTesterMixin, unittest.TestCase):
@@ -157,6 +177,21 @@ def test_gradient_checkpointing_is_applied(self):
         expected_set = {"HunyuanVideoTransformer3DModel"}
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
 
+    @require_torch_gpu
+    @require_torch_2
+    @is_torch_compile
+    @slow
+    def test_torch_compile_recompilation_and_graph_break(self):
+        torch._dynamo.reset()
+        init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()
+
+        model = self.model_class(**init_dict).to(torch_device)
+        model = torch.compile(model, fullgraph=True)
+
+        with torch._dynamo.config.patch(error_on_recompile=True), torch.no_grad():
+            _ = model(**inputs_dict)
+            _ = model(**inputs_dict)
+
 
 class HunyuanVideoImageToVideoTransformer3DTests(ModelTesterMixin, unittest.TestCase):
     model_class = HunyuanVideoTransformer3DModel
@@ -222,6 +257,21 @@ def test_output(self):
     def test_gradient_checkpointing_is_applied(self):
         expected_set = {"HunyuanVideoTransformer3DModel"}
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
+
+    @require_torch_gpu
+    @require_torch_2
+    @is_torch_compile
+    @slow
+    def test_torch_compile_recompilation_and_graph_break(self):
+        torch._dynamo.reset()
+        init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()
+
+        model = self.model_class(**init_dict).to(torch_device)
+        model = torch.compile(model, fullgraph=True)
+
+        with torch._dynamo.config.patch(error_on_recompile=True), torch.no_grad():
+            _ = model(**inputs_dict)
+            _ = model(**inputs_dict)
 
 
 class HunyuanVideoTokenReplaceImageToVideoTransformer3DTests(ModelTesterMixin, unittest.TestCase):
@@ -286,7 +336,22 @@ def prepare_init_args_and_inputs_for_common(self):
 
     def test_output(self):
         super().test_output(expected_output_shape=(1, *self.output_shape))
-
+    
     def test_gradient_checkpointing_is_applied(self):
         expected_set = {"HunyuanVideoTransformer3DModel"}
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
+
+    @require_torch_gpu
+    @require_torch_2
+    @is_torch_compile
+    @slow
+    def test_torch_compile_recompilation_and_graph_break(self):
+        torch._dynamo.reset()
+        init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()
+
+        model = self.model_class(**init_dict).to(torch_device)
+        model = torch.compile(model, fullgraph=True)
+
+        with torch._dynamo.config.patch(error_on_recompile=True), torch.no_grad():
+            _ = model(**inputs_dict)
+            _ = model(**inputs_dict)