新增 Torchvision 转换规则

paconvert/api_matcher.py

@@ -4792,3 +4792,129 @@ def generate_code(self, kwargs):
                     self.kwargs_to_str(kwargs_bin_edges),
                 )
         return code
+
+
+class DatasetFolderMatcher(BaseMatcher):
+    def generate_code(self, kwargs):
+        if "root" in kwargs:
+            kwargs["root"] = "str(Path({}))".format(kwargs["root"])
+        API_TEMPLATE = textwrap.dedent(
+            """
+            from pathlib import Path
+            {}({})
+            """
+        )
+        return API_TEMPLATE.format(self.get_paddle_api(), self.kwargs_to_str(kwargs))
+
+
+class ImageFolderMatcher(BaseMatcher):
+    def generate_code(self, kwargs):
+        if "root" in kwargs:
+            kwargs["root"] = "str(Path({}))".format(kwargs["root"])
+        API_TEMPLATE = textwrap.dedent(
+            """
+            from pathlib import Path
+            {}({})
+            """
+        )
+        return API_TEMPLATE.format(self.get_paddle_api(), self.kwargs_to_str(kwargs))
+
+
+class Cifar10Matcher(BaseMatcher):
+    def generate_code(self, kwargs):
+        if "root" in kwargs:
+            root = kwargs.pop("root")
+            data_file = "cifar-10-python.tar.gz"
+            kwargs["data_file"] = "str(Path({}) / '{}')".format(root, data_file)
+        if "train" in kwargs:
+            kwargs["mode"] = "'train' if {} else 'test'".format(kwargs.pop("train"))
+
+        API_TEMPLATE = textwrap.dedent(
+            """
+            from pathlib import Path
+            {}({})
+            """
+        )
+        return API_TEMPLATE.format(self.get_paddle_api(), self.kwargs_to_str(kwargs))
+
+
+class Cifar100Matcher(BaseMatcher):
+    def generate_code(self, kwargs):
+        if "root" in kwargs:
+            root = kwargs.pop("root")
+            data_file = "cifar-100-python.tar.gz"
+            kwargs["data_file"] = "str(Path({}) / '{}')".format(root, data_file)
+        if "train" in kwargs:
+            kwargs["mode"] = "'train' if {} else 'test'".format(kwargs.pop("train"))
+
+        API_TEMPLATE = textwrap.dedent(
+            """
+            from pathlib import Path
+            {}({})
+            """
+        )
+        return API_TEMPLATE.format(self.get_paddle_api(), self.kwargs_to_str(kwargs))
+
+
+class MNISTMatcher(BaseMatcher):
+    def generate_code(self, kwargs):
+        train = True
+        if "train" in kwargs:
+            train = kwargs.pop("train")
+            kwargs["mode"] = "'train' if {} else 'test'".format(train)
+        if "root" in kwargs:
+            root = kwargs.pop("root")
+            file_paths = {
+                "train_image": "MNIST/raw/train-images-idx3-ubyte.gz",
+                "train_label": "MNIST/raw/train-labels-idx1-ubyte.gz",
+                "test_image": "MNIST/raw/t10k-images-idx3-ubyte.gz",
+                "test_label": "MNIST/raw/t10k-labels-idx1-ubyte.gz",
+            }
+            kwargs["image_path"] = (
+                f"str(Path({root}) / '{file_paths['train_image']}') if {train} else "
+                f"str(Path({root}) / '{file_paths['test_image']}')"
+            )
+            kwargs["label_path"] = (
+                f"str(Path({root}) / '{file_paths['train_label']}') if {train} else "
+                f"str(Path({root}) / '{file_paths['test_label']}')"
+            )
+
+        API_TEMPLATE = textwrap.dedent(
+            """
+            from pathlib import Path
+            {}({})
+            """
+        )
+        return API_TEMPLATE.format(self.get_paddle_api(), self.kwargs_to_str(kwargs))
+
+
+class FashionMNISTMatcher(BaseMatcher):
+    def generate_code(self, kwargs):
+        train = True
+        if "train" in kwargs:
+            train = kwargs.pop("train")
+            kwargs["mode"] = "'train' if {} else 'test'".format(train)
+        if "root" in kwargs:
+            root = kwargs.pop("root")
+            file_paths = {
+                "train_image": "FashionMNIST/raw/train-images-idx3-ubyte.gz",
+                "train_label": "FashionMNIST/raw/train-labels-idx1-ubyte.gz",
+                "test_image": "FashionMNIST/raw/t10k-images-idx3-ubyte.gz",
+                "test_label": "FashionMNIST/raw/t10k-labels-idx1-ubyte.gz",
+            }
+            kwargs["image_path"] = (
+                f"str(Path({root}) / '{file_paths['train_image']}') if {train} else "
+                f"str(Path({root}) / '{file_paths['test_image']}')"
+            )
+            kwargs["label_path"] = (
+                f"str(Path({root}) / '{file_paths['train_label']}') if {train} else "
+                f"str(Path({root}) / '{file_paths['test_label']}')"
+            )
+
+        API_TEMPLATE = textwrap.dedent(
+            """
+            from pathlib import Path
+            {}({})
+            """
+        )
+        return API_TEMPLATE.format(self.get_paddle_api(), self.kwargs_to_str(kwargs))

tests/vision/__init__.py

@@ -0,0 +1,14 @@
+# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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.
+#

tests/vision/image_apibase.py

@@ -0,0 +1,62 @@
+# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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 numpy as np
+from apibase import APIBase
+from PIL import Image
+
+
+class ImageAPIBase(APIBase):
+    def compare(
+        self,
+        name,
+        pytorch_result,
+        paddle_result,
+        check_value=True,
+        check_dtype=True,
+        check_stop_gradient=True,
+        rtol=1.0e-6,
+        atol=0.0,
+    ):
+        """
+        Compare PIL Images for equality.
+        """
+        if isinstance(pytorch_result, Image.Image) and isinstance(
+            paddle_result, Image.Image
+        ):
+            pytorch_array = np.array(pytorch_result)
+            paddle_array = np.array(paddle_result)
+
+            assert (
+                pytorch_array.shape == paddle_array.shape
+            ), "API ({}): shape mismatch, torch shape is {}, paddle shape is {}".format(
+                name, pytorch_array.shape, paddle_array.shape
+            )
+
+            if check_value:
+                assert np.array_equal(
+                    pytorch_array, paddle_array
+                ), "API ({}): image data mismatch".format(name)
+            return
+
+        super().compare(
+            name,
+            pytorch_result,
+            paddle_result,
+            check_value,
+            check_dtype,
+            check_stop_gradient,
+            rtol,
+            atol,
+        )

tests/vision/test_CenterCrop.py

@@ -0,0 +1,93 @@
+# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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 textwrap
+
+from apibase import APIBase
+from vision.image_apibase import ImageAPIBase
+
+obj = APIBase("torchvision.transforms.CenterCrop")
+img_obj = ImageAPIBase("torchvision.transforms.CenterCrop")
+
+
+def test_case_1():
+    pytorch_code = textwrap.dedent(
+        """
+        import torch
+        import torchvision.transforms as transforms
+        img = torch.tensor([[[0.5, 0.5], [0.5, 0.5]],
+                             [[0.5, 0.5], [0.5, 0.5]],
+                             [[0.5, 0.5], [0.5, 0.5]]])
+        center_crop = transforms.CenterCrop((1, 1))
+        result = center_crop(img)
+        """
+    )
+    obj.run(pytorch_code, ["result"])
+
+
+def test_case_2():
+    pytorch_code = textwrap.dedent(
+        """
+        import torch
+        import torchvision.transforms as transforms
+        img = torch.tensor([[[0.1, 0.4], [0.7, 1.0]],
+                             [[0.2, 0.5], [0.8, 1.0]],
+                             [[0.3, 0.6], [0.9, 1.0]]])
+        center_crop = transforms.CenterCrop([1, 1])
+        result = center_crop(img)
+        """
+    )
+    obj.run(pytorch_code, ["result"])
+
+
+def test_case_3():
+    pytorch_code = textwrap.dedent(
+        """
+        from PIL import Image
+        import torchvision.transforms as transforms
+        img = Image.new('RGB', (4, 4), color=(100, 100, 100))
+        center_crop = transforms.CenterCrop(2)
+        result = center_crop(img)
+        """
+    )
+    img_obj.run(pytorch_code, ["result"])
+
+
+def test_case_4():
+    pytorch_code = textwrap.dedent(
+        """
+        import torch
+        import torchvision.transforms as transforms
+        img = torch.tensor([[[0.1, 0.2, 0.3, 0.4],
+                              [0.5, 0.6, 0.7, 0.8]],
+                             [[0.9, 1.0, 1.1, 1.2],
+                              [1.3, 1.4, 1.5, 1.6]]])
+        center_crop = transforms.CenterCrop((2, 2))
+        result = center_crop(img)
+        """
+    )
+    obj.run(pytorch_code, ["result"])
+
+
+def test_case_5():
+    pytorch_code = textwrap.dedent(
+        """
+        from PIL import Image
+        import torchvision.transforms as transforms
+        img = Image.new('RGB', (3, 3), color=(50, 100, 150))
+        center_crop = transforms.CenterCrop((2, 2))
+        result = center_crop(img)
+        """
+    )
+    img_obj.run(pytorch_code, ["result"])