[CI] Refactor Chronoedit, PRX, EasyAnimate, Ovis transformer tests

tests/models/transformers/test_models_transformer_chronoedit.py

@@ -0,0 +1,101 @@
+# coding=utf-8
+# Copyright 2025 HuggingFace Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+
+from diffusers import ChronoEditTransformer3DModel
+from diffusers.utils.torch_utils import randn_tensor
+
+from ...testing_utils import enable_full_determinism, torch_device
+from ..testing_utils import (
+    BaseModelTesterConfig,
+    ModelTesterMixin,
+    TrainingTesterMixin,
+)
+
+
+enable_full_determinism()
+
+
+class ChronoEditTransformerTesterConfig(BaseModelTesterConfig):
+    @property
+    def model_class(self):
+        return ChronoEditTransformer3DModel
+
+    @property
+    def main_input_name(self) -> str:
+        return "hidden_states"
+
+    @property
+    def output_shape(self) -> tuple:
+        return (16, 8, 8)
+
+    @property
+    def input_shape(self) -> tuple:
+        return (16, 8, 8)
+
+    @property
+    def generator(self):
+        return torch.Generator("cpu").manual_seed(0)
+
+    def get_init_dict(self) -> dict:
+        return {
+            "patch_size": (1, 2, 2),
+            "num_attention_heads": 2,
+            "attention_head_dim": 8,
+            "in_channels": 16,
+            "out_channels": 16,
+            "text_dim": 32,
+            "freq_dim": 16,
+            "ffn_dim": 32,
+            "num_layers": 2,
+            "cross_attn_norm": True,
+            "qk_norm": "rms_norm_across_heads",
+            "eps": 1e-06,
+            "image_dim": None,
+            "added_kv_proj_dim": None,
+            "rope_max_seq_len": 64,
+            "pos_embed_seq_len": None,
+            "rope_temporal_skip_len": 8,
+        }
+
+    def get_dummy_inputs(self, batch_size: int = 1) -> dict[str, torch.Tensor]:
+        num_channels = 16
+        num_frames = 2
+        height = 8
+        width = 8
+        embedding_dim = 32
+        sequence_length = 12
+
+        return {
+            "hidden_states": randn_tensor(
+                (batch_size, num_channels, num_frames, height, width), generator=self.generator, device=torch_device
+            ),
+            "timestep": torch.randint(0, 1000, size=(batch_size,), generator=self.generator).to(torch_device),
+            "encoder_hidden_states": randn_tensor(
+                (batch_size, sequence_length, embedding_dim), generator=self.generator, device=torch_device
+            ),
+            "encoder_hidden_states_image": None,
+        }
+
+
+class TestChronoEditTransformer(ChronoEditTransformerTesterConfig, ModelTesterMixin):
+    pass
+
+
+class TestChronoEditTransformerTraining(ChronoEditTransformerTesterConfig, TrainingTesterMixin):
+    def test_gradient_checkpointing_is_applied(self):
+        expected_set = {"ChronoEditTransformer3DModel"}
+        super().test_gradient_checkpointing_is_applied(expected_set=expected_set)

tests/models/transformers/test_models_transformer_easyanimate.py

@@ -13,58 +13,45 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import unittest
-
 import torch
 
 from diffusers import EasyAnimateTransformer3DModel
+from diffusers.utils.torch_utils import randn_tensor
 
 from ...testing_utils import enable_full_determinism, torch_device
-from ..test_modeling_common import ModelTesterMixin
+from ..testing_utils import (
+    BaseModelTesterConfig,
+    ModelTesterMixin,
+    TrainingTesterMixin,
+)
 
 
 enable_full_determinism()
 
 
-class EasyAnimateTransformerTests(ModelTesterMixin, unittest.TestCase):
-    model_class = EasyAnimateTransformer3DModel
-    main_input_name = "hidden_states"
-    uses_custom_attn_processor = True
-
+class EasyAnimateTransformerTesterConfig(BaseModelTesterConfig):
     @property
-    def dummy_input(self):
-        batch_size = 2
-        num_channels = 4
-        num_frames = 2
-        height = 16
-        width = 16
-        embedding_dim = 16
-        sequence_length = 16
-
-        hidden_states = torch.randn((batch_size, num_channels, num_frames, height, width)).to(torch_device)
-        encoder_hidden_states = torch.randn((batch_size, sequence_length, embedding_dim)).to(torch_device)
-        timestep = torch.randint(0, 1000, size=(batch_size,)).to(torch_device)
+    def model_class(self):
+        return EasyAnimateTransformer3DModel
 
-        return {
-            "hidden_states": hidden_states,
-            "timestep": timestep,
-            "timestep_cond": None,
-            "encoder_hidden_states": encoder_hidden_states,
-            "encoder_hidden_states_t5": None,
-            "inpaint_latents": None,
-            "control_latents": None,
-        }
+    @property
+    def main_input_name(self) -> str:
+        return "hidden_states"
 
     @property
-    def input_shape(self):
+    def output_shape(self) -> tuple:
         return (4, 2, 16, 16)
 
     @property
-    def output_shape(self):
+    def input_shape(self) -> tuple:
         return (4, 2, 16, 16)
 
-    def prepare_init_args_and_inputs_for_common(self):
-        init_dict = {
+    @property
+    def generator(self):
+        return torch.Generator("cpu").manual_seed(0)
+
+    def get_init_dict(self) -> dict:
+        return {
             "attention_head_dim": 16,
             "num_attention_heads": 2,
             "in_channels": 4,
@@ -79,9 +66,35 @@ def prepare_init_args_and_inputs_for_common(self):
             "time_position_encoding_type": "3d_rope",
             "timestep_activation_fn": "silu",
         }
-        inputs_dict = self.dummy_input
-        return init_dict, inputs_dict
 
+    def get_dummy_inputs(self, batch_size: int = 2) -> dict[str, torch.Tensor]:
+        num_channels = 4
+        num_frames = 2
+        height = 16
+        width = 16
+        embedding_dim = 16
+        sequence_length = 16
+
+        return {
+            "hidden_states": randn_tensor(
+                (batch_size, num_channels, num_frames, height, width), generator=self.generator, device=torch_device
+            ),
+            "timestep": torch.randint(0, 1000, size=(batch_size,), generator=self.generator).to(torch_device),
+            "timestep_cond": None,
+            "encoder_hidden_states": randn_tensor(
+                (batch_size, sequence_length, embedding_dim), generator=self.generator, device=torch_device
+            ),
+            "encoder_hidden_states_t5": None,
+            "inpaint_latents": None,
+            "control_latents": None,
+        }
+
+
+class TestEasyAnimateTransformer(EasyAnimateTransformerTesterConfig, ModelTesterMixin):
+    pass
+
+
+class TestEasyAnimateTransformerTraining(EasyAnimateTransformerTesterConfig, TrainingTesterMixin):
     def test_gradient_checkpointing_is_applied(self):
         expected_set = {"EasyAnimateTransformer3DModel"}
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)

tests/models/transformers/test_models_transformer_hunyuan_video.py

@@ -22,7 +22,7 @@
     enable_full_determinism,
     torch_device,
 )
-from ..test_modeling_common import ModelTesterMixin, TorchCompileTesterMixin
+from ..test_modeling_common import ModelTesterMixin
 
 
 enable_full_determinism()
@@ -93,13 +93,6 @@ def test_gradient_checkpointing_is_applied(self):
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
 
 
-class HunyuanTransformerCompileTests(TorchCompileTesterMixin, unittest.TestCase):
-    model_class = HunyuanVideoTransformer3DModel
-
-    def prepare_init_args_and_inputs_for_common(self):
-        return HunyuanVideoTransformer3DTests().prepare_init_args_and_inputs_for_common()
-
-
 class HunyuanSkyreelsImageToVideoTransformer3DTests(ModelTesterMixin, unittest.TestCase):
     model_class = HunyuanVideoTransformer3DModel
     main_input_name = "hidden_states"
@@ -168,13 +161,6 @@ def test_gradient_checkpointing_is_applied(self):
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
 
 
-class HunyuanSkyreelsImageToVideoCompileTests(TorchCompileTesterMixin, unittest.TestCase):
-    model_class = HunyuanVideoTransformer3DModel
-
-    def prepare_init_args_and_inputs_for_common(self):
-        return HunyuanSkyreelsImageToVideoTransformer3DTests().prepare_init_args_and_inputs_for_common()
-
-
 class HunyuanVideoImageToVideoTransformer3DTests(ModelTesterMixin, unittest.TestCase):
     model_class = HunyuanVideoTransformer3DModel
     main_input_name = "hidden_states"
@@ -241,13 +227,6 @@ def test_gradient_checkpointing_is_applied(self):
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
 
 
-class HunyuanImageToVideoCompileTests(TorchCompileTesterMixin, unittest.TestCase):
-    model_class = HunyuanVideoTransformer3DModel
-
-    def prepare_init_args_and_inputs_for_common(self):
-        return HunyuanVideoImageToVideoTransformer3DTests().prepare_init_args_and_inputs_for_common()
-
-
 class HunyuanVideoTokenReplaceImageToVideoTransformer3DTests(ModelTesterMixin, unittest.TestCase):
     model_class = HunyuanVideoTransformer3DModel
     main_input_name = "hidden_states"
@@ -314,10 +293,3 @@ def test_output(self):
     def test_gradient_checkpointing_is_applied(self):
         expected_set = {"HunyuanVideoTransformer3DModel"}
         super().test_gradient_checkpointing_is_applied(expected_set=expected_set)
-
-
-class HunyuanVideoTokenReplaceCompileTests(TorchCompileTesterMixin, unittest.TestCase):
-    model_class = HunyuanVideoTransformer3DModel
-
-    def prepare_init_args_and_inputs_for_common(self):
-        return HunyuanVideoTokenReplaceImageToVideoTransformer3DTests().prepare_init_args_and_inputs_for_common()

tests/models/transformers/test_models_transformer_ovis_image.py

@@ -0,0 +1,97 @@
+# coding=utf-8
+# Copyright 2025 HuggingFace Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+
+from diffusers import OvisImageTransformer2DModel
+from diffusers.utils.torch_utils import randn_tensor
+
+from ...testing_utils import enable_full_determinism, torch_device
+from ..testing_utils import (
+    BaseModelTesterConfig,
+    ModelTesterMixin,
+    TrainingTesterMixin,
+)
+
+
+enable_full_determinism()
+
+
+class OvisImageTransformerTesterConfig(BaseModelTesterConfig):
+    @property
+    def model_class(self):
+        return OvisImageTransformer2DModel
+
+    @property
+    def main_input_name(self) -> str:
+        return "hidden_states"
+
+    @property
+    def output_shape(self) -> tuple:
+        return (16, 4)
+
+    @property
+    def input_shape(self) -> tuple:
+        return (16, 4)
+
+    @property
+    def generator(self):
+        return torch.Generator("cpu").manual_seed(0)
+
+    def get_init_dict(self) -> dict:
+        return {
+            "patch_size": 1,
+            "in_channels": 4,
+            "out_channels": 4,
+            "num_layers": 1,
+            "num_single_layers": 1,
+            "attention_head_dim": 16,
+            "num_attention_heads": 2,
+            "joint_attention_dim": 32,
+            "axes_dims_rope": (4, 4, 8),
+        }
+
+    def get_dummy_inputs(self, batch_size: int = 1) -> dict[str, torch.Tensor]:
+        num_latent_channels = 4
+        num_image_channels = 3
+        height = width = 4
+        sequence_length = 48
+        embedding_dim = 32
+
+        return {
+            "hidden_states": randn_tensor(
+                (batch_size, height * width, num_latent_channels), generator=self.generator, device=torch_device
+            ),
+            "encoder_hidden_states": randn_tensor(
+                (batch_size, sequence_length, embedding_dim), generator=self.generator, device=torch_device
+            ),
+            "img_ids": randn_tensor(
+                (height * width, num_image_channels), generator=self.generator, device=torch_device
+            ),
+            "txt_ids": randn_tensor(
+                (sequence_length, num_image_channels), generator=self.generator, device=torch_device
+            ),
+            "timestep": torch.tensor([1.0]).to(torch_device).expand(batch_size),
+        }
+
+
+class TestOvisImageTransformer(OvisImageTransformerTesterConfig, ModelTesterMixin):
+    pass
+
+
+class TestOvisImageTransformerTraining(OvisImageTransformerTesterConfig, TrainingTesterMixin):
+    def test_gradient_checkpointing_is_applied(self):
+        expected_set = {"OvisImageTransformer2DModel"}
+        super().test_gradient_checkpointing_is_applied(expected_set=expected_set)