Moved nih_cxr14_re configs to data

pyproject.toml

@@ -471,10 +471,10 @@ mc_ch_in_pc_rgb = "ptbench.data.mc_ch_in_pc.rgb"
 # (with radiological signs)
 mc_ch_in_pc_rs = "ptbench.configs.datasets.mc_ch_in_pc_RS.default"
 # NIH CXR14 (relabeled)
-nih_cxr14 = "ptbench.configs.datasets.nih_cxr14_re.default"
-nih_cxr14_cm = "ptbench.configs.datasets.nih_cxr14_re.cardiomegaly"
+nih_cxr14 = "ptbench.data.nih_cxr14_re.default"
+nih_cxr14_cm = "ptbench.data.nih_cxr14_re.cardiomegaly"
 # NIH CXR14 / PadChest aggregated dataset
-nih_cxr14_pc_idiap = "ptbench.configs.datasets.nih_cxr14_re_pc.idiap"
+nih_cxr14_pc_idiap = "ptbench.data.nih_cxr14_re_pc.idiap"
 # PadChest
 padchest_idiap = "ptbench.data.padchest.idiap"
 padchest_tb_idiap = "ptbench.data.padchest.tb_idiap"

src/ptbench/configs/datasets/nih_cxr14_re/__init__.py

-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-
-def _maker(protocol, size=512):
-    import torchvision.transforms as transforms
-
-    from ....data.nih_cxr14_re import dataset as raw
-    from .. import make_dataset as mk
-
-    # ImageNet normalization
-    normalize = transforms.Normalize(
-        mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
-    )
-
-    return mk(
-        [raw.subsets(protocol)],
-        [transforms.Resize((size, size))],
-        [transforms.RandomHorizontalFlip()],
-        [transforms.ToTensor(), normalize],
-    )

src/ptbench/data/nih_cxr14_re/__init__.py

@@ -72,9 +72,27 @@ def _loader(context, sample):
     return make_delayed(sample, _raw_data_loader)
 
 
-dataset = JSONDataset(
+json_dataset = JSONDataset(
     protocols=_protocols,
     fieldnames=("data", "label"),
     loader=_loader,
 )
 """NIH CXR14 (relabeled) dataset object."""
+
+
+def _maker(protocol, size=512):
+    import torchvision.transforms as transforms
+
+    from .. import make_dataset
+
+    # ImageNet normalization
+    normalize = transforms.Normalize(
+        mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
+    )
+
+    return make_dataset(
+        [json_dataset.subsets(protocol)],
+        [transforms.Resize((size, size))],
+        [transforms.RandomHorizontalFlip()],
+        [transforms.ToTensor(), normalize],
+    )

src/ptbench/configs/datasets/nih_cxr14_re/cardiomegaly.py → src/ptbench/data/nih_cxr14_re/cardiomegaly.py

@@ -11,6 +11,38 @@ First 40 images with cardiomegaly.
 * See :py:mod:`ptbench.data.nih_cxr14_re` for dataset details
 """
 
+from clapper.logging import setup
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule
 from . import _maker
 
-dataset = _maker("cardiomegaly")
+logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+
+
+class Fold0Module(BaseDataModule):
+    def __init__(
+        self,
+        train_batch_size=1,
+        predict_batch_size=1,
+        drop_incomplete_batch=False,
+        multiproc_kwargs=None,
+    ):
+        super().__init__(
+            train_batch_size=train_batch_size,
+            predict_batch_size=predict_batch_size,
+            drop_incomplete_batch=drop_incomplete_batch,
+            multiproc_kwargs=multiproc_kwargs,
+        )
+
+    def setup(self, stage: str):
+        self.dataset = _maker("cardiomegaly")
+        (
+            self.train_dataset,
+            self.validation_dataset,
+            self.extra_validation_datasets,
+            self.predict_dataset,
+        ) = return_subsets(self.dataset)
+
+
+datamodule = Fold0Module

src/ptbench/configs/datasets/nih_cxr14_re/default.py → src/ptbench/data/nih_cxr14_re/default.py

@@ -10,6 +10,38 @@ protocol)
 * See :py:mod:`ptbench.data.nih_cxr14_re` for dataset details
 """
 
+from clapper.logging import setup
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule
 from . import _maker
 
-dataset = _maker("default")
+logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+
+
+class DefaultModule(BaseDataModule):
+    def __init__(
+        self,
+        train_batch_size=1,
+        predict_batch_size=1,
+        drop_incomplete_batch=False,
+        multiproc_kwargs=None,
+    ):
+        super().__init__(
+            train_batch_size=train_batch_size,
+            predict_batch_size=predict_batch_size,
+            drop_incomplete_batch=drop_incomplete_batch,
+            multiproc_kwargs=multiproc_kwargs,
+        )
+
+    def setup(self, stage: str):
+        self.dataset = _maker("default")
+        (
+            self.train_dataset,
+            self.validation_dataset,
+            self.extra_validation_datasets,
+            self.predict_dataset,
+        ) = return_subsets(self.dataset)
+
+
+datamodule = DefaultModule