quadra 2.2.5__py3-none-any.whl → 2.3.0__py3-none-any.whl

quadra/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "2.2.5"
+__version__ = "2.3.0"
 
 
 def get_version():

quadra/callbacks/anomalib.py

@@ -64,7 +64,7 @@ class Visualizer:
         self.figure.subplots_adjust(right=0.9)
 
         axes = self.axis if len(self.images) > 1 else [self.axis]
-        for axis, image_dict in zip(axes, self.images):
+        for axis, image_dict in zip(axes, self.images, strict=False):
             axis.axes.xaxis.set_visible(False)
             axis.axes.yaxis.set_visible(False)
             axis.imshow(image_dict["image"], image_dict["color_map"], vmin=0, vmax=255)
@@ -201,6 +201,7 @@ class VisualizerCallback(Callback):
                 outputs["label"],
                 outputs["pred_labels"],
                 outputs["pred_scores"],
+                strict=False,
             )
         ):
             denormalized_image = Denormalize()(image.cpu())
@@ -256,7 +257,7 @@ class VisualizerCallback(Callback):
             visualizer.close()
 
             if self.plot_raw_outputs:
-                for raw_output, raw_name in zip([heatmap, vis_img], ["heatmap", "segmentation"]):
+                for raw_output, raw_name in zip([heatmap, vis_img], ["heatmap", "segmentation"], strict=False):
                     current_raw_output = raw_output
                     if raw_name == "segmentation":
                         current_raw_output = (raw_output * 255).astype(np.uint8)

quadra/callbacks/lightning.py

@@ -79,6 +79,8 @@ def _scale_batch_size(
         new_size = _run_power_scaling(trainer, init_val, batch_arg_name, max_trials, params)
     elif mode == "binsearch":
         new_size = _run_binary_scaling(trainer, init_val, batch_arg_name, max_trials, params)
+    else:
+        raise ValueError(f"Unknown mode {mode}")
 
     garbage_collection_cuda()
 

quadra/configs/callbacks/all.yaml

@@ -30,3 +30,16 @@ progress_bar:
 lightning_trainer_setup:
   _target_: quadra.callbacks.lightning.LightningTrainerBaseSetup
   log_every_n_steps: 1
+
+batch_size_finder:
+  _target_: quadra.callbacks.lightning.BatchSizeFinder
+  mode: power
+  steps_per_trial: 3
+  init_val: 2
+  max_trials: 5 # Max 64
+  batch_arg_name: batch_size
+  disable: false
+  find_train_batch_size: true
+  find_validation_batch_size: false
+  find_test_batch_size: false
+  find_predict_batch_size: false

quadra/configs/callbacks/default.yaml

@@ -9,9 +9,6 @@ model_checkpoint:
   filename: "epoch_{epoch:03d}"
   auto_insert_metric_name: False
 
-log_gradients:
-  _target_: quadra.callbacks.mlflow.LogGradients
-  norm: 2
 lr_monitor:
   _target_: pytorch_lightning.callbacks.LearningRateMonitor
   logging_interval: "epoch"

quadra/configs/callbacks/default_anomalib.yaml

@@ -44,9 +44,6 @@ upload_ckpts_as_artifact:
   upload_best_only: true
   delete_after_upload: true
   upload: false
-log_gradients:
-  _target_: quadra.callbacks.mlflow.LogGradients
-  norm: 2
 lr_monitor:
   _target_: pytorch_lightning.callbacks.LearningRateMonitor
   logging_interval: "epoch"

quadra/datamodules/base.py

@@ -7,7 +7,7 @@ import pickle as pkl
 import typing
 from collections.abc import Callable, Iterable, Sequence
 from functools import wraps
-from typing import Any, Literal, Union, cast
+from typing import Any, Literal, cast
 
 import albumentations
 import numpy as np
@@ -20,8 +20,8 @@ from tqdm import tqdm
 from quadra.utils import utils
 
 log = utils.get_logger(__name__)
-TrainDataset = Union[torch.utils.data.Dataset, Sequence[torch.utils.data.Dataset]]
-ValDataset = Union[torch.utils.data.Dataset, Sequence[torch.utils.data.Dataset]]
+TrainDataset = torch.utils.data.Dataset | Sequence[torch.utils.data.Dataset]
+ValDataset = torch.utils.data.Dataset | Sequence[torch.utils.data.Dataset]
 TestDataset = torch.utils.data.Dataset
 
 
@@ -260,7 +260,7 @@ class BaseDataModule(LightningDataModule, metaclass=DecorateParentMethod):
             return
 
         # TODO: We need to find a way to annotate the columns of data.
-        paths_and_hash_length = zip(self.data["samples"], [self.hash_size] * len(self.data))
+        paths_and_hash_length = zip(self.data["samples"], [self.hash_size] * len(self.data), strict=False)
 
         with mp.Pool(min(8, mp.cpu_count() - 1)) as pool:
             self.data["hash"] = list(
@@ -355,7 +355,7 @@ class BaseDataModule(LightningDataModule, metaclass=DecorateParentMethod):
             raise ValueError("`n_aug_to_take` is not set. Cannot load augmented samples.")
         aug_samples = []
         aug_labels = []
-        for sample, label in zip(samples, targets):
+        for sample, label in zip(samples, targets, strict=False):
             aug_samples.append(sample)
             aug_labels.append(label)
             final_sample = sample

quadra/datamodules/classification.py

@@ -243,7 +243,7 @@ class ClassificationDataModule(BaseDataModule):
                 samples_test, targets_test = self._read_split(self.test_split_file)
                 if not self.train_split_file:
                     samples_train, targets_train = [], []
-                    for sample, target in zip(all_samples, all_targets):
+                    for sample, target in zip(all_samples, all_targets, strict=False):
                         if sample not in samples_test:
                             samples_train.append(sample)
                             targets_train.append(target)
@@ -251,7 +251,7 @@ class ClassificationDataModule(BaseDataModule):
                 samples_train, targets_train = self._read_split(self.train_split_file)
                 if not self.test_split_file:
                     samples_test, targets_test = [], []
-                    for sample, target in zip(all_samples, all_targets):
+                    for sample, target in zip(all_samples, all_targets, strict=False):
                         if sample not in samples_train:
                             samples_test.append(sample)
                             targets_test.append(target)

quadra/datamodules/segmentation.py

@@ -187,7 +187,7 @@ class SegmentationDataModule(BaseDataModule):
             samples_test, targets_test, masks_test = self._read_split(self.test_split_file)
             if not self.train_split_file:
                 samples_train, targets_train, masks_train = [], [], []
-                for sample, target, mask in zip(all_samples, all_targets, all_masks):
+                for sample, target, mask in zip(all_samples, all_targets, all_masks, strict=False):
                     if sample not in samples_test:
                         samples_train.append(sample)
                         targets_train.append(target)
@@ -197,7 +197,7 @@ class SegmentationDataModule(BaseDataModule):
             samples_train, targets_train, masks_train = self._read_split(self.train_split_file)
             if not self.test_split_file:
                 samples_test, targets_test, masks_test = [], [], []
-                for sample, target, mask in zip(all_samples, all_targets, all_masks):
+                for sample, target, mask in zip(all_samples, all_targets, all_masks, strict=False):
                     if sample not in samples_train:
                         samples_test.append(sample)
                         targets_test.append(target)
@@ -549,7 +549,7 @@ class SegmentationMulticlassDataModule(BaseDataModule):
                 samples_and_masks_test,
                 targets_test,
             ) = iterative_train_test_split(
-                np.expand_dims(np.array(list(zip(all_samples, all_masks))), 1),
+                np.expand_dims(np.array(list(zip(all_samples, all_masks, strict=False))), 1),
                 np.array(all_targets),
                 test_size=self.test_size,
             )
@@ -561,7 +561,7 @@ class SegmentationMulticlassDataModule(BaseDataModule):
             samples_test, targets_test, masks_test = self._read_split(self.test_split_file)
             if not self.train_split_file:
                 samples_train, targets_train, masks_train = [], [], []
-                for sample, target, mask in zip(all_samples, all_targets, all_masks):
+                for sample, target, mask in zip(all_samples, all_targets, all_masks, strict=False):
                     if sample not in samples_test:
                         samples_train.append(sample)
                         targets_train.append(target)
@@ -571,7 +571,7 @@ class SegmentationMulticlassDataModule(BaseDataModule):
             samples_train, targets_train, masks_train = self._read_split(self.train_split_file)
             if not self.test_split_file:
                 samples_test, targets_test, masks_test = [], [], []
-                for sample, target, mask in zip(all_samples, all_targets, all_masks):
+                for sample, target, mask in zip(all_samples, all_targets, all_masks, strict=False):
                     if sample not in samples_train:
                         samples_test.append(sample)
                         targets_test.append(target)
@@ -583,7 +583,7 @@ class SegmentationMulticlassDataModule(BaseDataModule):
                 raise ValueError("Validation split file is specified but no train or test split file is specified.")
         else:
             samples_and_masks_train, targets_train, samples_and_masks_val, targets_val = iterative_train_test_split(
-                np.expand_dims(np.array(list(zip(samples_train, masks_train))), 1),
+                np.expand_dims(np.array(list(zip(samples_train, masks_train, strict=False))), 1),
                 np.array(targets_train),
                 test_size=self.val_size,
             )

quadra/datasets/anomaly.py

@@ -220,7 +220,7 @@ class AnomalyDataset(Dataset):
             if not os.path.exists(valid_area_mask):
                 raise RuntimeError(f"Valid area mask {valid_area_mask} does not exist.")
 
-            self.valid_area_mask = cv2.imread(valid_area_mask, 0) > 0  # type: ignore[operator]
+            self.valid_area_mask = cv2.imread(valid_area_mask, 0) > 0
 
     def __len__(self) -> int:
         """Get length of the dataset."""
@@ -265,7 +265,7 @@ class AnomalyDataset(Dataset):
                 if label_index == 0:
                     mask = np.zeros(shape=original_image_shape[:2])
                 elif os.path.isfile(mask_path):
-                    mask = cv2.imread(mask_path, flags=0) / 255.0  # type: ignore[operator]
+                    mask = cv2.imread(mask_path, flags=0) / 255.0
                 else:
                     # We need ones in the mask to compute correctly at least image level f1 score
                     mask = np.ones(shape=original_image_shape[:2])

quadra/datasets/classification.py

@@ -50,9 +50,9 @@ class ImageClassificationListDataset(Dataset):
         allow_missing_label: bool | None = False,
     ):
         super().__init__()
-        assert len(samples) == len(
-            targets
-        ), f"Samples ({len(samples)}) and targets ({len(targets)}) must have the same length"
+        assert len(samples) == len(targets), (
+            f"Samples ({len(samples)}) and targets ({len(targets)}) must have the same length"
+        )
         # Setting the ROI
         self.roi = roi
 
@@ -201,9 +201,9 @@ class MultilabelClassificationDataset(torch.utils.data.Dataset):
         rgb: bool = True,
     ):
         super().__init__()
-        assert len(samples) == len(
-            targets
-        ), f"Samples ({len(samples)}) and targets ({len(targets)}) must have the same length"
+        assert len(samples) == len(targets), (
+            f"Samples ({len(samples)}) and targets ({len(targets)}) must have the same length"
+        )
 
         # Data
         self.x = samples
@@ -215,7 +215,7 @@ class MultilabelClassificationDataset(torch.utils.data.Dataset):
             class_to_idx = {c: i for i, c in enumerate(range(unique_targets))}
         self.class_to_idx = class_to_idx
         self.idx_to_class = {v: k for k, v in class_to_idx.items()}
-        self.samples = list(zip(self.x, self.y))
+        self.samples = list(zip(self.x, self.y, strict=False))
         self.rgb = rgb
         self.transform = transform
 

quadra/datasets/patch.py

@@ -58,7 +58,7 @@ class PatchSklearnClassificationTrainDataset(Dataset):
 
             cls, counts = np.unique(targets_array, return_counts=True)
             max_count = np.max(counts)
-            for cl, count in zip(cls, counts):
+            for cl, count in zip(cls, counts, strict=False):
                 idx_to_pick = list(np.where(targets_array == cl)[0])
 
                 if count < max_count:

quadra/metrics/segmentation.py

@@ -171,7 +171,7 @@ def segmentation_props(
         # Add dummy Dices so LSA is unique and i can compute FP and FN
         dice_mat = _pad_to_shape(dice_mat, (max_dim, max_dim), 1)
         lsa = linear_sum_assignment(dice_mat, maximize=False)
-        for row, col in zip(lsa[0], lsa[1]):
+        for row, col in zip(lsa[0], lsa[1], strict=False):
             # More preds than GTs --> False Positive
             if row < n_labels_pred and col >= n_labels_mask:
                 min_row = pred_bbox[row][0]

quadra/models/base.py

@@ -76,7 +76,7 @@ class ModelSignatureWrapper(nn.Module):
 
         if isinstance(self.instance.forward, torch.ScriptMethod):
             # Handle torchscript backbones
-            for i, argument in enumerate(self.instance.forward.schema.arguments):
+            for i, argument in enumerate(self.instance.forward.schema.arguments):  # type: ignore[attr-defined]
                 if i < (len(args) + 1):  # +1 for self
                     continue
 

quadra/models/evaluation.py

@@ -209,7 +209,7 @@ class ONNXEvaluationModel(BaseEvaluationModel):
 
         onnx_inputs: dict[str, np.ndarray | torch.Tensor] = {}
 
-        for onnx_input, current_input in zip(self.model.get_inputs(), inputs):
+        for onnx_input, current_input in zip(self.model.get_inputs(), inputs, strict=False):
             if isinstance(current_input, torch.Tensor):
                 onnx_inputs[onnx_input.name] = current_input
                 use_pytorch = True

quadra/modules/base.py

@@ -7,6 +7,7 @@ import pytorch_lightning as pl
 import sklearn
 import torch
 import torchmetrics
+from pytorch_lightning.utilities.types import OptimizerLRScheduler
 from sklearn.linear_model import LogisticRegression
 from torch import nn
 from torch.optim import Optimizer
@@ -48,7 +49,7 @@ class BaseLightningModule(pl.LightningModule):
         """
         return self.model(x)
 
-    def configure_optimizers(self) -> tuple[list[Any], list[dict[str, Any]]]:
+    def configure_optimizers(self) -> OptimizerLRScheduler:
         """Get default optimizer if not passed a value.
 
         Returns:
@@ -68,7 +69,7 @@ class BaseLightningModule(pl.LightningModule):
             "monitor": "val_loss",
             "strict": False,
         }
-        return [self.optimizer], [lr_scheduler_conf]
+        return [self.optimizer], [lr_scheduler_conf]  # type: ignore[return-value]
 
     # pylint: disable=unused-argument
     def optimizer_zero_grad(self, epoch, batch_idx, optimizer, optimizer_idx: int = 0):

quadra/modules/ssl/byol.py

@@ -110,6 +110,7 @@ class BYOL(SSLModule):
             for student_ps, teacher_ps in zip(
                 list(self.model.parameters()) + list(self.student_projection_mlp.parameters()),
                 list(self.teacher.parameters()) + list(self.teacher_projection_mlp.parameters()),
+                strict=False,
             ):
                 teacher_ps.data = teacher_ps.data * teacher_momentum + (1 - teacher_momentum) * student_ps.data
 

quadra/tasks/anomaly.py

@@ -161,7 +161,7 @@ class AnomalibDetection(Generic[AnomalyDataModuleT], LightningTask[AnomalyDataMo
         all_output_flatten: dict[str, torch.Tensor | list] = {}
 
         for key in all_output[0]:
-            if type(all_output[0][key]) == torch.Tensor:
+            if isinstance(all_output[0][key], torch.Tensor):
                 tensor_gatherer = torch.cat([x[key] for x in all_output])
                 all_output_flatten[key] = tensor_gatherer
             else:
@@ -205,13 +205,15 @@ class AnomalibDetection(Generic[AnomalyDataModuleT], LightningTask[AnomalyDataMo
             class_to_idx.pop("false_defect")
 
         anomaly_scores = all_output_flatten["pred_scores"]
+
+        exportable_anomaly_scores: list[Any] | np.ndarray
         if isinstance(anomaly_scores, torch.Tensor):
             exportable_anomaly_scores = anomaly_scores.cpu().numpy()
         else:
             exportable_anomaly_scores = anomaly_scores
 
         # Zip the lists together to create rows for the CSV file
-        rows = zip(image_paths, pred_labels, gt_labels, exportable_anomaly_scores)
+        rows = zip(image_paths, pred_labels, gt_labels, exportable_anomaly_scores, strict=False)
         # Specify the CSV file name
         csv_file = "test_predictions.csv"
         # Write the data to the CSV file
@@ -483,7 +485,7 @@ class AnomalibEvaluation(Evaluation[AnomalyDataModule]):
 
         if hasattr(self.datamodule, "valid_area_mask") and self.datamodule.valid_area_mask is not None:
             mask_area = cv2.imread(self.datamodule.valid_area_mask, 0)
-            mask_area = (mask_area > 0).astype(np.uint8)  # type: ignore[operator]
+            mask_area = (mask_area > 0).astype(np.uint8)
 
         if hasattr(self.datamodule, "crop_area") and self.datamodule.crop_area is not None:
             crop_area = self.datamodule.crop_area
@@ -499,12 +501,13 @@ class AnomalibEvaluation(Evaluation[AnomalyDataModule]):
                 self.metadata["image_labels"],
                 anomaly_scores,
                 anomaly_maps,
+                strict=False,
             ),
             total=len(self.metadata["image_paths"]),
         ):
             img = cv2.imread(img_path, 0)
             if mask_area is not None:
-                img = img * mask_area  # type: ignore[operator]
+                img = img * mask_area
 
             if crop_area is not None:
                 img = img[crop_area[1] : crop_area[3], crop_area[0] : crop_area[2]]

quadra/tasks/base.py

@@ -382,15 +382,19 @@ class Evaluation(Generic[DataModuleT], Task[DataModuleT]):
             # We assume that each input size has the same height and width
             if input_size[1] != self.config.transforms.input_height:
                 log.warning(
-                    f"Input height of the model ({input_size[1]}) is different from the one specified "
-                    + f"in the config ({self.config.transforms.input_height}). Fixing the config."
+                    "Input height of the model (%s) is different from the one specified "
+                    + "in the config (%s). Fixing the config.",
+                    input_size[1],
+                    self.config.transforms.input_height,
                 )
                 self.config.transforms.input_height = input_size[1]
 
             if input_size[2] != self.config.transforms.input_width:
                 log.warning(
-                    f"Input width of the model ({input_size[2]}) is different from the one specified "
-                    + f"in the config ({self.config.transforms.input_width}). Fixing the config."
+                    "Input width of the model (%s) is different from the one specified "
+                    + "in the config (%s). Fixing the config.",
+                    input_size[2],
+                    self.config.transforms.input_width,
                 )
                 self.config.transforms.input_width = input_size[2]
 

quadra/tasks/classification.py

@@ -623,7 +623,9 @@ class SklearnClassification(Generic[SklearnClassificationDataModuleT], Task[Skle
             all_labels = all_labels[sorted_indices]
 
         # cycle over all train/test split
-        for train_dataloader, test_dataloader in zip(self.train_dataloader_list, self.test_dataloader_list):
+        for train_dataloader, test_dataloader in zip(
+            self.train_dataloader_list, self.test_dataloader_list, strict=False
+        ):
             # Reinit classifier
             self.model = self.config.model
             self.trainer.change_classifier(self.model)
@@ -685,7 +687,7 @@ class SklearnClassification(Generic[SklearnClassificationDataModuleT], Task[Skle
             dl: PyTorch dataloader
             feature_extractor: PyTorch backbone
         """
-        if isinstance(feature_extractor, (TorchEvaluationModel, TorchscriptEvaluationModel)):
+        if isinstance(feature_extractor, TorchEvaluationModel | TorchscriptEvaluationModel):
             # TODO: I'm not sure torchinfo supports torchscript models
             # If we are working with torch based evaluation models we need to extract the model
             feature_extractor = feature_extractor.model
@@ -1202,6 +1204,8 @@ class ClassificationEvaluation(Evaluation[ClassificationDataModuleT]):
         probabilities = [max(item) for sublist in probabilities for item in sublist]
         if self.datamodule.class_to_idx is not None:
             idx_to_class = {v: k for k, v in self.datamodule.class_to_idx.items()}
+        else:
+            idx_to_class = None
 
         _, pd_cm, test_accuracy = get_results(
             test_labels=image_labels,

quadra/tasks/patch.py

@@ -301,7 +301,7 @@ class PatchSklearnTestClassification(Evaluation[PatchSklearnClassificationDataMo
             "test_results": None,
             "test_labels": None,
         }
-        self.class_to_skip: list[str] = []
+        self.class_to_skip: list[str] | None = []
         self.reconstruction_results: dict[str, Any]
         self.return_polygon: bool = True
 

quadra/tasks/segmentation.py

@@ -92,8 +92,10 @@ class Segmentation(Generic[SegmentationDataModuleT], LightningTask[SegmentationD
             len(self.datamodule.idx_to_class) + 1
         ):
             log.warning(
-                f"Number of classes in the model ({module_config.model.num_classes}) does not match the number of "
-                + f"classes in the datamodule ({len(self.datamodule.idx_to_class)}). Updating the model..."
+                "Number of classes in the model (%s) does not match the number of "
+                + "classes in the datamodule (%d). Updating the model...",
+                module_config.model.num_classes,
+                len(self.datamodule.idx_to_class),
             )
             module_config.model.num_classes = len(self.datamodule.idx_to_class) + 1
 
@@ -341,7 +343,7 @@ class SegmentationAnalysisEvaluation(SegmentationEvaluation):
         if self.datamodule.test_dataset_available:
             stages.append("test")
             dataloaders.append(self.datamodule.test_dataloader())
-        for stage, dataloader in zip(stages, dataloaders):
+        for stage, dataloader in zip(stages, dataloaders, strict=False):
             log.info("Running inference on %s set with batch size: %d", stage, dataloader.batch_size)
             image_list, mask_list, mask_pred_list, label_list = [], [], [], []
             for batch in dataloader:
@@ -369,10 +371,10 @@ class SegmentationAnalysisEvaluation(SegmentationEvaluation):
 
         for stage, output in self.test_output.items():
             image_mean = OmegaConf.to_container(self.config.transforms.mean)
-            if not isinstance(image_mean, list) or any(not isinstance(x, (int, float)) for x in image_mean):
+            if not isinstance(image_mean, list) or any(not isinstance(x, int | float) for x in image_mean):
                 raise ValueError("Image mean is not a list of float or integer values, please check your config")
             image_std = OmegaConf.to_container(self.config.transforms.std)
-            if not isinstance(image_std, list) or any(not isinstance(x, (int, float)) for x in image_std):
+            if not isinstance(image_std, list) or any(not isinstance(x, int | float) for x in image_std):
                 raise ValueError("Image std is not a list of float or integer values, please check your config")
             reports = create_mask_report(
                 stage=stage,

quadra/tasks/ssl.py

@@ -468,8 +468,7 @@ class EmbeddingVisualization(Task):
         self.report_folder = report_folder
         if self.model_path is None:
             raise ValueError(
-                "Model path cannot be found!, please specify it in the config or pass it as an argument for"
-                " evaluation"
+                "Model path cannot be found!, please specify it in the config or pass it as an argument for evaluation"
             )
         self.embeddings_path = os.path.join(self.model_path, self.report_folder)
         if not os.path.exists(self.embeddings_path):
@@ -547,7 +546,7 @@ class EmbeddingVisualization(Task):
                 im = interpolate(im, self.embedding_image_size)
 
             images.append(im.cpu())
-            metadata.extend(zip(targets, class_names, file_paths))
+            metadata.extend(zip(targets, class_names, file_paths, strict=False))
             counter += len(im)
         images = torch.cat(images, dim=0)
         embeddings = torch.cat(embeddings, dim=0)

quadra/utils/classification.py

@@ -46,12 +46,10 @@ def get_file_condition(
         if any(fil in root for fil in exclude_filter):
             return False
 
-    if include_filter is not None and (
-        not any(fil in file_name for fil in include_filter) and not any(fil in root for fil in include_filter)
-    ):
-        return False
-
-    return True
+    return not (
+        include_filter is not None
+        and (not any(fil in file_name for fil in include_filter) and not any(fil in root for fil in include_filter))
+    )
 
 
 def natural_key(string_):
@@ -130,7 +128,7 @@ def find_images_and_targets(
         sorted_labels = sorted(unique_labels, key=natural_key)
         class_to_idx = {str(c): idx for idx, c in enumerate(sorted_labels)}
 
-    images_and_targets = [(f, l) for f, l in zip(filenames, labels) if l in class_to_idx]
+    images_and_targets = [(f, l) for f, l in zip(filenames, labels, strict=False) if l in class_to_idx]
 
     if sort:
         images_and_targets = sorted(images_and_targets, key=lambda k: natural_key(k[0]))
@@ -210,7 +208,7 @@ def find_test_image(
             file_samples.append(sample_path)
 
         test_split = [os.path.join(folder, sample.strip()) for sample in file_samples]
-        labels = [t for s, t in zip(filenames, labels) if s in file_samples]
+        labels = [t for s, t in zip(filenames, labels, strict=False) if s in file_samples]
         filenames = [s for s in filenames if s in file_samples]
         log.info("Selected %d images using test_split_file for the test", len(filenames))
         if len(filenames) != len(file_samples):
@@ -353,7 +351,7 @@ def get_split(
 
     cl, counts = np.unique(targets, return_counts=True)
 
-    for num, _cl in zip(counts, cl):
+    for num, _cl in zip(counts, cl, strict=False):
         if num == 1:
             to_remove = np.where(np.array(targets) == _cl)[0][0]
             samples = np.delete(np.array(samples), to_remove)
@@ -378,7 +376,7 @@ def get_split(
             file_samples.append(sample_path)
 
         train_split = [os.path.join(image_dir, sample.strip()) for sample in file_samples]
-        targets = np.array([t for s, t in zip(samples, targets) if s in file_samples])
+        targets = np.array([t for s, t in zip(samples, targets, strict=False) if s in file_samples])
         samples = np.array([s for s in samples if s in file_samples])
 
     if limit_training_data is not None:

quadra/utils/evaluation.py

@@ -4,6 +4,7 @@ import os
 from ast import literal_eval
 from collections.abc import Callable
 from functools import wraps
+from typing import Any
 
 import matplotlib.pyplot as plt
 import numpy as np
@@ -123,7 +124,7 @@ def calculate_mask_based_metrics(
     th_thresh_preds = (th_preds > threshold).float().cpu()
     thresh_preds = th_thresh_preds.squeeze(0).numpy()
     dice_scores = metric(th_thresh_preds, th_masks, reduction=None).numpy()
-    result = {}
+    result: dict[str, Any] = {}
     if multilabel:
         if n_classes is None:
             raise ValueError("n_classes arg shouldn't be None when multilabel is True")
@@ -167,7 +168,7 @@ def calculate_mask_based_metrics(
         "Accuracy": [],
     }
     for idx, (image, pred, mask, thresh_pred, dice_score) in enumerate(
-        zip(images, preds, masks, thresh_preds, dice_scores)
+        zip(images, preds, masks, thresh_preds, dice_scores, strict=False)
     ):
         if np.sum(mask) == 0:
             good_dice.append(dice_score)
@@ -261,6 +262,7 @@ def create_mask_report(
     th_labels = output["label"]
     n_classes = th_preds.shape[1]
     # TODO: Apply sigmoid is a wrong name now
+    # TODO: Apply sigmoid false is untested
     if apply_sigmoid:
         if n_classes == 1:
             th_preds = torch.nn.Sigmoid()(th_preds)
@@ -271,6 +273,13 @@ def create_mask_report(
             # Compute labels from the given masks since by default they are all 0
             th_labels = th_masks.max(dim=2)[0].max(dim=2)[0].squeeze(dim=1)
             show_orj_predictions = False
+    elif n_classes == 1:
+        th_thresh_preds = (th_preds > threshold).float()
+    else:
+        th_thresh_preds = torch.argmax(th_preds, dim=1).float().unsqueeze(1)
+        # Compute labels from the given masks since by default they are all 0
+        th_labels = th_masks.max(dim=2)[0].max(dim=2)[0].squeeze(dim=1)
+        show_orj_predictions = False
 
     mean = np.asarray(mean)
     std = np.asarray(std)
@@ -303,7 +312,7 @@ def create_mask_report(
     non_zero_score_idx = sorted_idx[~binary_labels]
     zero_score_idx = sorted_idx[binary_labels]
     file_paths = []
-    for name, current_score_idx in zip(["good", "bad"], [zero_score_idx, non_zero_score_idx]):
+    for name, current_score_idx in zip(["good", "bad"], [zero_score_idx, non_zero_score_idx], strict=False):
         if len(current_score_idx) == 0:
             continue
 

quadra/utils/export.py

@@ -45,11 +45,11 @@ def generate_torch_inputs(
     """
     inp = None
 
-    if isinstance(input_shapes, (ListConfig, DictConfig)):
+    if isinstance(input_shapes, ListConfig | DictConfig):
         input_shapes = OmegaConf.to_container(input_shapes)
 
     if isinstance(input_shapes, list):
-        if any(isinstance(inp, (Sequence, dict)) for inp in input_shapes):
+        if any(isinstance(inp, Sequence | dict) for inp in input_shapes):
             return [generate_torch_inputs(inp, device, half_precision, dtype) for inp in input_shapes]
 
         # Base case
@@ -59,7 +59,7 @@ def generate_torch_inputs(
         return {k: generate_torch_inputs(v, device, half_precision, dtype) for k, v in input_shapes.items()}
 
     if isinstance(input_shapes, tuple):
-        if any(isinstance(inp, (Sequence, dict)) for inp in input_shapes):
+        if any(isinstance(inp, Sequence | dict) for inp in input_shapes):
             # The tuple contains a list, tuple or dict
             return tuple(generate_torch_inputs(inp, device, half_precision, dtype) for inp in input_shapes)
 
@@ -381,7 +381,7 @@ def _safe_export_half_precision_onnx(
             with open(os.devnull, "w") as f, contextlib.redirect_stdout(f):
                 # This function prints a lot of information that is not useful for the user
                 model_fp16 = auto_convert_mixed_precision(
-                    model_fp32, test_data, rtol=0.01, atol=0.001, keep_io_types=False
+                    model_fp32, test_data, rtol=0.01, atol=5e-3, keep_io_types=False
                 )
             onnx.save(model_fp16, export_model_path)
 

quadra/utils/mlflow.py

@@ -11,6 +11,7 @@ except ImportError:
 from collections.abc import Sequence
 from typing import Any
 
+import numpy as np
 import torch
 from pytorch_lightning import Trainer
 from pytorch_lightning.loggers import MLFlowLogger
@@ -45,6 +46,7 @@ def infer_signature_input(input_tensor: Any) -> Any:
     Raises:
         ValueError: If the input type is not supported or when nested dicts or sequences are encountered.
     """
+    signature: dict[str, Any] | np.ndarray
     if isinstance(input_tensor, Sequence):
         # Mlflow currently does not support sequence outputs, so we use a dict instead
         signature = {}

quadra/utils/models.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 import math
 import warnings
 from collections.abc import Callable
-from typing import Union, cast
+from typing import cast
 
 import numpy as np
 import timm
@@ -114,7 +114,7 @@ def get_feature(
             labels: input_labels
             grayscale_cams: Gradcam output maps, None if gradcam arg is False
     """
-    if isinstance(feature_extractor, (TorchEvaluationModel, TorchscriptEvaluationModel)):
+    if isinstance(feature_extractor, TorchEvaluationModel | TorchscriptEvaluationModel):
         # If we are working with torch based evaluation models we need to extract the model
         feature_extractor = feature_extractor.model
     elif isinstance(feature_extractor, ONNXEvaluationModel):
@@ -160,9 +160,7 @@ def get_feature(
                 x1 = x1.to(feature_extractor.device).to(feature_extractor.model_dtype)
 
             if gradcam:
-                y_hat = cast(
-                    Union[list[torch.Tensor], tuple[torch.Tensor], torch.Tensor], feature_extractor(x1).detach()
-                )
+                y_hat = cast(list[torch.Tensor] | tuple[torch.Tensor] | torch.Tensor, feature_extractor(x1).detach())
                 # mypy can't detect that gradcam is true only if we have a features_extractor
                 if is_vision_transformer(feature_extractor.features_extractor):  # type: ignore[union-attr]
                     grayscale_cam_low_res = grad_rollout(
@@ -177,10 +175,10 @@ def get_feature(
                 feature_extractor.zero_grad(set_to_none=True)  # type: ignore[union-attr]
             else:
                 with torch.no_grad():
-                    y_hat = cast(Union[list[torch.Tensor], tuple[torch.Tensor], torch.Tensor], feature_extractor(x1))
+                    y_hat = cast(list[torch.Tensor] | tuple[torch.Tensor] | torch.Tensor, feature_extractor(x1))
                 grayscale_cams = None
 
-            if isinstance(y_hat, (list, tuple)):
+            if isinstance(y_hat, list | tuple):
                 y_hat = y_hat[0].cpu()
             else:
                 y_hat = y_hat.cpu()

quadra/utils/patch/dataset.py

@@ -566,7 +566,7 @@ def generate_patch_dataset(
             num_workers=num_workers,
         )
 
-    for phase, split_dict in zip(["val", "test"], [val_data_dictionary, test_data_dictionary]):
+    for phase, split_dict in zip(["val", "test"], [val_data_dictionary, test_data_dictionary], strict=False):
         if len(split_dict) > 0:
             log.info("Generating %s set", phase)
             generate_patch_sliding_window_dataset(
@@ -908,9 +908,9 @@ def extract_patches(
             patches = np.concatenate([patches, extra_patches_h], axis=0)
 
     # If this is not true there's some strange case I didn't take into account
-    assert (
-        patches.shape[0] == patch_num_h and patches.shape[1] == patch_num_w
-    ), f"Patch shape {patches.shape} does not match the expected shape {patch_number}"
+    assert patches.shape[0] == patch_num_h and patches.shape[1] == patch_num_w, (
+        f"Patch shape {patches.shape} does not match the expected shape {patch_number}"
+    )
 
     return patches
 
@@ -1059,11 +1059,12 @@ def create_h5(
         h = img.shape[0]
         w = img.shape[1]
 
+        mask: np.ndarray
         if item["mask"] is None:
-            mask = np.zeros([h, w])
+            mask = np.zeros([h, w], dtype=np.uint8)
         else:
             # this works even if item["mask"] is already an absolute path
-            mask = cv2.imread(os.path.join(output_folder, item["mask"]), 0)  # type: ignore[assignment]
+            mask = cv2.imread(os.path.join(output_folder, item["mask"]), 0)
 
         if patch_size is not None:
             patch_height = patch_size[1]

quadra/utils/patch/metrics.py

@@ -98,9 +98,9 @@ def compute_patch_metrics(
     if (patch_h is not None and patch_w is not None) and (patch_num_h is not None and patch_num_w is not None):
         raise ValueError("Either number of patches or patch size is required for reconstruction")
 
-    assert (patch_h is not None and patch_w is not None) or (
-        patch_num_h is not None and patch_num_w is not None
-    ), "Either number of patches or patch size is required for reconstruction"
+    assert (patch_h is not None and patch_w is not None) or (patch_num_h is not None and patch_num_w is not None), (
+        "Either number of patches or patch size is required for reconstruction"
+    )
 
     if patch_h is not None and patch_w is not None and patch_num_h is not None and patch_num_w is not None:
         warnings.warn(
@@ -191,7 +191,7 @@ def compute_patch_metrics(
             if annotated_good is not None:
                 gt_img[np.isin(gt_img, annotated_good)] = 0
 
-            gt_img_binary = (gt_img > 0).astype(bool)  # type: ignore[operator]
+            gt_img_binary = (gt_img > 0).astype(bool)
             regions_pred = label(output_mask).astype(np.uint8)
 
             for k in range(1, regions_pred.max() + 1):
@@ -203,8 +203,11 @@ def compute_patch_metrics(
             output_mask = (output_mask > 0).astype(np.uint8)
             gt_img = label(gt_img)
 
-            for i in range(1, gt_img.max() + 1):  # type: ignore[union-attr]
-                region = (gt_img == i).astype(bool)  # type: ignore[union-attr]
+            if gt_img is None:
+                raise RuntimeError("Ground truth mask is None after label and it should not be")
+
+            for i in range(1, gt_img.max() + 1):
+                region = (gt_img == i).astype(bool)
                 if np.sum(np.bitwise_and(region, output_mask)) == 0:
                     false_region_good += 1
                 else:

quadra/utils/patch/visualization.py

@@ -69,13 +69,13 @@ def plot_patch_reconstruction(
             points = [[item["x"], item["y"]] for item in region["points"]]
             c_label = region["label"]
 
-            out = cv2.drawContours(
+            out = cv2.drawContours(  # type: ignore[call-overload]
                 out,
                 np.array([points], np.int32),
                 -1,
                 class_to_idx[c_label],
                 thickness=cv2.FILLED,
-            )  # type: ignore[call-overload]
+            )
     else:
         out = reconstruction["prediction"]
 

quadra/utils/utils.py

@@ -438,7 +438,7 @@ def flatten_list(input_list: Iterable[Any]) -> Iterator[Any]:
         The iterator over the flattend list
     """
     for v in input_list:
-        if isinstance(v, Iterable) and not isinstance(v, (str, bytes)):
+        if isinstance(v, Iterable) and not isinstance(v, str | bytes):
             yield from flatten_list(v)
         else:
             yield v

quadra/utils/validator.py

@@ -72,9 +72,7 @@ def check_all_arguments(callable_variable: str, configuration_arguments: list[st
     """
     for argument in configuration_arguments:
         if argument not in argument_names:
-            error_string = (
-                f"`{argument}` is not a valid argument passed " f"from configuration to `{callable_variable}`."
-            )
+            error_string = f"`{argument}` is not a valid argument passed from configuration to `{callable_variable}`."
             closest_match = difflib.get_close_matches(argument, argument_names, n=1, cutoff=0.5)
             if len(closest_match) > 0:
                 error_string += f" Did you mean `{closest_match[0]}`?"

quadra/utils/visualization.py

@@ -46,7 +46,7 @@ class UnNormalize:
             new_t = tensor.detach().clone()
         else:
             new_t = tensor
-        for t, m, s in zip(new_t, self.mean, self.std):
+        for t, m, s in zip(new_t, self.mean, self.std, strict=False):
             t.mul_(s).add_(m)
             # The normalize code -> t.sub_(m).div_(s)
         return new_t
@@ -82,7 +82,7 @@ def create_grid_figure(
             ax[i][j].get_xaxis().set_ticks([])
             ax[i][j].get_yaxis().set_ticks([])
     if row_names is not None:
-        for ax, name in zip(ax[:, 0], row_names):  # noqa: B020
+        for ax, name in zip(ax[:, 0], row_names, strict=False):  # noqa: B020
             ax.set_ylabel(name, rotation=90)
 
     plt.tight_layout()
@@ -98,12 +98,12 @@ def create_visualization_dataset(dataset: torch.utils.data.Dataset):
         """Handle different types of transforms."""
         if isinstance(transforms, albumentations.BaseCompose):
             transforms.transforms = convert_transforms(transforms.transforms)
-        if isinstance(transforms, (list, ListConfig, TransformsSeqType)):
+        if isinstance(transforms, list | ListConfig | TransformsSeqType):
             transforms = [convert_transforms(t) for t in transforms]
-        if isinstance(transforms, (dict, DictConfig)):
+        if isinstance(transforms, dict | DictConfig):
             for tname, t in transforms.items():
                 transforms[tname] = convert_transforms(t)
-        if isinstance(transforms, (Normalize, ToTensorV2)):
+        if isinstance(transforms, Normalize | ToTensorV2):
             return NoOp(p=1)
         return transforms
 
@@ -362,6 +362,9 @@ def plot_classification_results(
                         test_label = idx_to_class[test_labels[i]]
                     except Exception:
                         test_label = test_labels[i]
+                else:
+                    pred_label = pred_labels[i]
+                    test_label = test_labels[i]
 
                 ax.axis("off")
                 ax.set_title(f"True: {str(test_label)}\nPred {str(pred_label)}")

quadra/utils/vit_explainability.py

@@ -153,7 +153,7 @@ def grad_rollout(
     """
     result = torch.eye(attentions[0].size(-1))
     with torch.no_grad():
-        for attention, grad in zip(attentions, gradients):
+        for attention, grad in zip(attentions, gradients, strict=False):
             weights = grad
             attention_heads_fused = torch.mean((attention * weights), dim=1)
             attention_heads_fused[attention_heads_fused < 0] = 0

{quadra-2.2.5.dist-info → quadra-2.3.0.dist-info}/METADATA

@@ -1,19 +1,18 @@
 Metadata-Version: 2.1
 Name: quadra
-Version: 2.2.5
+Version: 2.3.0
 Summary: Deep Learning experiment orchestration library
 Home-page: https://orobix.github.io/quadra
 License: Apache-2.0
 Keywords: deep learning,experiment,lightning,hydra-core
 Author: Federico Belotti
 Author-email: federico.belotti@orobix.com
-Requires-Python: >=3.9,<3.11
+Requires-Python: >=3.10,<3.11
 Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Education
 Classifier: Intended Audience :: Science/Research
 Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Software Development
@@ -21,7 +20,7 @@ Classifier: Topic :: Software Development :: Libraries
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Provides-Extra: onnx
 Requires-Dist: albumentations (>=1.3,<1.4)
-Requires-Dist: anomalib-orobix (==0.7.0.dev143)
+Requires-Dist: anomalib-orobix (==0.7.0.dev150)
 Requires-Dist: boto3 (>=1.26,<1.27)
 Requires-Dist: grad-cam-orobix (==1.5.3.dev001)
 Requires-Dist: h5py (>=3.8,<3.9)
@@ -32,17 +31,18 @@ Requires-Dist: label_studio_converter (>=0.0,<0.1)
 Requires-Dist: matplotlib (>=3.6,<3.7)
 Requires-Dist: minio (>=7.1,<7.2)
 Requires-Dist: mlflow-skinny (>=2.3.1,<3.0.0)
+Requires-Dist: numpy (<2)
 Requires-Dist: nvitop (>=0.11,<0.12)
 Requires-Dist: onnx (==1.15.0) ; extra == "onnx"
 Requires-Dist: onnxconverter-common (>=1.14.0,<2.0.0) ; extra == "onnx"
-Requires-Dist: onnxruntime_gpu (==1.17.0) ; extra == "onnx"
+Requires-Dist: onnxruntime_gpu (==1.20.0) ; extra == "onnx"
 Requires-Dist: onnxsim (==0.4.28) ; extra == "onnx"
 Requires-Dist: opencv_python_headless (>=4.7.0,<4.8.0)
 Requires-Dist: pandas (<2.0)
-Requires-Dist: pillow (>=9.3,<9.4)
+Requires-Dist: pillow (>=10,<11)
 Requires-Dist: pydantic (==1.10.10)
 Requires-Dist: python_dotenv (>=0.21,<0.22)
-Requires-Dist: pytorch_lightning (>=2.1,<2.2)
+Requires-Dist: pytorch_lightning (>=2.4,<2.5)
 Requires-Dist: rich (>=13.2,<13.3)
 Requires-Dist: scikit_learn (>=1.2,<1.3)
 Requires-Dist: scikit_multilearn (>=0.2,<0.3)
@@ -50,11 +50,11 @@ Requires-Dist: seaborn (>=0.12,<0.13)
 Requires-Dist: segmentation_models_pytorch-orobix (==0.3.3.dev1)
 Requires-Dist: tensorboard (>=2.11,<2.12)
 Requires-Dist: timm (==0.9.12)
-Requires-Dist: torch (==2.1.2)
+Requires-Dist: torch (==2.4.1)
 Requires-Dist: torchinfo (>=1.8,<1.9)
 Requires-Dist: torchmetrics (>=0.10,<0.11)
 Requires-Dist: torchsummary (>=1.5,<1.6)
-Requires-Dist: torchvision (>=0.16,<0.17)
+Requires-Dist: torchvision (>=0.19,<0.20)
 Requires-Dist: tripy (>=1.0,<1.1)
 Requires-Dist: typing_extensions (==4.11.0) ; python_version < "3.10"
 Requires-Dist: xxhash (>=3.2,<3.3)

{quadra-2.2.5.dist-info → quadra-2.3.0.dist-info}/RECORD

@@ -1,7 +1,7 @@
-quadra/__init__.py,sha256=wi1UBd-2e2ycj46ChXS4E9NfP8aKtbcXQohx6bZlmsg,112
+quadra/__init__.py,sha256=s2xCezYlfSL46r5iADHZuB2zg-0_EgoN-FNv4N4Kw48,112
 quadra/callbacks/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-quadra/callbacks/anomalib.py,sha256=Z0Xx3M9UEMYGY0BpSsqemgQSQ6jAr4SrvlbDg_oG9C8,11913
-quadra/callbacks/lightning.py,sha256=1OTM6fB7qBVLPWNBAZJOb3B00q0kAxMWkPjTEn9YgF0,20182
+quadra/callbacks/anomalib.py,sha256=WLBEGhZA9HoP4Yh9UbbC2GzDOKYTkvU9EY1lkZcV7Fs,11971
+quadra/callbacks/lightning.py,sha256=mRWvBCrxHry_MfvAHy9H2jRYCcyUCpurXS8B5TpoScs,20241
 quadra/callbacks/mlflow.py,sha256=4LKjrgbRCHP5dOCoDpF7J25gaBgABa0Rof-EA61Iqug,10129
 quadra/callbacks/scheduler.py,sha256=zrglcTUvMO236VchQFtCSlA-XXhc6a3HVWX0uDVQoyc,2656
 quadra/configs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -32,9 +32,9 @@ quadra/configs/backbone/vit16_base.yaml,sha256=LWMAx3uwKQClDds1zUMzdNercHgO3kanV
 quadra/configs/backbone/vit16_small.yaml,sha256=8VkdiFCDugoSwz3tEbTNYp54P2eHElC8dR9Tea2Qir8,198
 quadra/configs/backbone/vit16_tiny.yaml,sha256=iHk8eUeieg-MJF4YA7m7p5WiTz5KT_rmNmYA5LLjdaw,197
 quadra/configs/backbone/xcit_tiny_24_p8_224.yaml,sha256=GnlQpUcahn_kZpj109kCC_5X4HP9-BSlgXRkiQJTlGU,196
-quadra/configs/callbacks/all.yaml,sha256=KYEDjqAf7lC5gpU8uUjlfpETIiDaWq5AlnQN-OCOTZU,1113
-quadra/configs/callbacks/default.yaml,sha256=sJBXHcrgycjGiz2lfSn54s1c1fbuED2QLKMaJdos078,1254
-quadra/configs/callbacks/default_anomalib.yaml,sha256=spzxnmnl5RRWOxzWSIZDmtb3cz9bGAErZEO-PIAFoqE,2330
+quadra/configs/callbacks/all.yaml,sha256=LZx8d0apwv9t0KlKqzFCrYo2NpuNBgcUnHhG_kud5fQ,1437
+quadra/configs/callbacks/default.yaml,sha256=ZFPU1bm36hJsxI-85uiJx7TpX7qWkR8dibBKWtES4Yc,1180
+quadra/configs/callbacks/default_anomalib.yaml,sha256=FjjSj6HgMvH18MV4AKR4Ew0pOerPefj9eUGinUkODLE,2256
 quadra/configs/config.yaml,sha256=IULhqUF8Z7Cqr5Xx41EGj8dwtPQWSRiZ5jwzNg0Rjwk,686
 quadra/configs/core/default.yaml,sha256=IfgjKHXuOknq3CKvKKBqMPfiqmZSUOnc80Q2Jkbm7go,239
 quadra/configs/datamodule/base/anomaly.yaml,sha256=CILLAoQHunrT4BN0ynOzizTX29k-7B9vDVIZUYm-cBU,377
@@ -187,20 +187,20 @@ quadra/configs/transforms/dino.yaml,sha256=NtEbtJPHYkR9DOBPwXR33uvrEdqv8WNyqRlXr
 quadra/configs/transforms/linear_eval.yaml,sha256=fXmJBEwTWQ-QBMNV0mSG9wcrj31YGIV_czcRDczc1ss,488
 quadra/datamodules/__init__.py,sha256=y00iX2YAy6CJzPstKSBNq8_1YsYTRr_sCvqaL-WI7Z8,636
 quadra/datamodules/anomaly.py,sha256=_3FZNSwdMj-ECXlPQDslswtaMn0F1EgzA0q0UH-UgFY,6670
-quadra/datamodules/base.py,sha256=vDKyoZMMzKUn9dSU1hNiCW4Tt7G6xFTxuZn6c8CQxIc,14001
-quadra/datamodules/classification.py,sha256=UfszA_14TmS-9xQLdVLVJAjUT2PG1KLzfOX2w9mFKzg,41594
+quadra/datamodules/base.py,sha256=QGkJ8Lq6hznHvaXjD8mhJhrinrs4ZFlZD3-B5cLU0cQ,14010
+quadra/datamodules/classification.py,sha256=VwQd-zhzJuLgq5Kg1niOY4pnRbO7Sk4B77dWiTFv4do,41622
 quadra/datamodules/generic/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 quadra/datamodules/generic/imagenette.py,sha256=3hOb-GmvnKx_hqrSqRcAcf22PjtCQ3CY_-5mlaZSTIM,5564
 quadra/datamodules/generic/mnist.py,sha256=j4xWEWQb1utW3yyozgHD1tP0kOAtLpRsgeIBZ1cIiP0,3425
 quadra/datamodules/generic/mvtec.py,sha256=3Ib8JyY1Eg7wbPL2dXw22YCoy_gsitksofFShLQ9Itw,2700
 quadra/datamodules/generic/oxford_pet.py,sha256=tumWy9TBThvVQZ2JOyghosWJEEsYjyXN6pZMJ9C5dBY,6822
 quadra/datamodules/patch.py,sha256=y7leDt1MyVg0LnqKgWCZ0i6cuVln10fiG4X8EFbl-_Q,7789
-quadra/datamodules/segmentation.py,sha256=eWS_v7H4ea9h05KW1nMpLMJjCGVZEnzBU-vuKlVAdeY,28920
+quadra/datamodules/segmentation.py,sha256=hhfOs7QoYslHYfWfnAgZzSusj2tus8k-h7SBqGNVT8E,29004
 quadra/datamodules/ssl.py,sha256=U63FCdcRJjx4K0RZzkKJfvYJhFpvWTnlBBCtXirn_F4,5709
 quadra/datasets/__init__.py,sha256=nVpqp2ffQ6omqCMB3r1ajcUGgUad0eSkDt-kNWDGblU,669
-quadra/datasets/anomaly.py,sha256=Guxb39aZkne5Qp_pEhjPACdzYSkLdLbu5AUXP0Gf1aE,11944
-quadra/datasets/classification.py,sha256=e-hNFh76aFnvbtkQLq1ljJ9Q3a6ymOmFziNjNK8VM4U,7512
-quadra/datasets/patch.py,sha256=bloUJS_qzzSfVqZaFpQd0uJMhRKRCfEwnSqCJaVtyZQ,4789
+quadra/datasets/anomaly.py,sha256=4rCd2-frgMH3RfQYVFYn5ZXxTKbPOk8GwE-BZIiLwFY,11892
+quadra/datasets/classification.py,sha256=ISKcY2PwD3HNv1JPPbDIJRJWJmu3KR3hlx3HUxlXYpE,7530
+quadra/datasets/patch.py,sha256=imNJONPoREivSZ-6WqYO2zE80PDEr-oCm3rdJuKlWz0,4803
 quadra/datasets/segmentation.py,sha256=cDs45eRh_IBSLB0K5xDos-D4KySRQN64BzaPKGBF7OI,9056
 quadra/datasets/ssl.py,sha256=bbGWM-mQvr5xqXtmanr6HKC1Hgq42asu7wx2prb5NVo,3925
 quadra/losses/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -219,21 +219,21 @@ quadra/losses/ssl/simsiam.py,sha256=uCCbqU9aYMwNa3re0qkeEK5Iz7Hxi0jAcEc-sCWZ8fc,
 quadra/losses/ssl/vicreg.py,sha256=ANvhYJz6iPv1A-OBXgBSrZrDG-1VmPtK1IZDtyFqNHE,2427
 quadra/main.py,sha256=6ZYKytVvCzQjgP_0QA6-3ICzVppsbRgPjF-csLKv85o,1407
 quadra/metrics/__init__.py,sha256=HsTK1gxsjp8_MYgA5caa4OK8sXLqtK_tt9wYyjtFnOc,79
-quadra/metrics/segmentation.py,sha256=jkQzCUz0ibkEoF95-uL-4wwxybs5SAs2VHiT5ZGckso,9451
+quadra/metrics/segmentation.py,sha256=tVRYEyMiwD0RJ7NtoGRoSbwb8sAKoVmvzEhV6-3iQT4,9465
 quadra/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-quadra/models/base.py,sha256=nYbZ1JRmhgcORZhAJoV6d2_L5C5yds_X-PQAFIDoHWk,5514
+quadra/models/base.py,sha256=QqMRQWqIsUuUHeInqMHZv3wv7Xeqz-zKe4cAQeqaN3M,5544
 quadra/models/classification/__init__.py,sha256=c03CGDyvipXgU7ybSosOaeTl1aM1ge6TqMUgMiTpQtA,243
 quadra/models/classification/backbones.py,sha256=haHNPC-XZ8Jj1i47cfUj8JHy_I-rins-nNfccrPBffo,6281
 quadra/models/classification/base.py,sha256=w-mDPQPtIrNclxjqsve5BTmNhNgnWGh7uJfE5HaTFPA,2996
-quadra/models/evaluation.py,sha256=dEazagyUvlnnEq16IdM25qAdDXZR-BUhApDF_YUdO-E,10695
+quadra/models/evaluation.py,sha256=LQg2K6PDIKK0ZnkP4pHfRNnKO4WeaROoYoNFA3Bctg0,10709
 quadra/modules/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 quadra/modules/backbone.py,sha256=xiZBqgzr1S45GX9mydl29TFuahLDaHrU7Fy73LGIyGI,909
-quadra/modules/base.py,sha256=fS_kbOAfEihHFfiTcyuyDXTzkmkJr2ivtAMiZqjOUog,10303
+quadra/modules/base.py,sha256=y96PSFJeo4gswVj3a6uNnoirg-dMgS0MsYDN51fQQ9A,10382
 quadra/modules/classification/__init__.py,sha256=6keltBhC1yzgbNttBuykNYJAUMyOrY-HDNgGZGfI93I,141
 quadra/modules/classification/base.py,sha256=QdHtHY2tF_qh2wU01Oo0TWjh9CTqa46tyF4VgcLd__M,11937
 quadra/modules/ssl/__init__.py,sha256=oeUoGHrsESZ0595-JxPxURBP124jtNfrITbVovBpANA,302
 quadra/modules/ssl/barlowtwins.py,sha256=iW6f7ADSEkbs7z-88x680204-Ez-iF1Yd2SdQzcLpRY,1884
-quadra/modules/ssl/byol.py,sha256=8sviU7MS9MXDbS9Ogu9-qJCZrkTI2pPxgEwNvc7EqIo,7084
+quadra/modules/ssl/byol.py,sha256=3UhUr72kpI2lM9JtVPqrTcpTo60NsAHNu3SIwD5_RrI,7114
 quadra/modules/ssl/common.py,sha256=nQMsYEu4PUueMq0KNe898h3wGS2RVQBN0NCpYnMyRqI,9898
 quadra/modules/ssl/dino.py,sha256=Xs4wRYvvxeLuHtOW5Gf-xaqAvT97cIuOG6PlYduPDm4,7300
 quadra/modules/ssl/hyperspherical.py,sha256=yEY0WvYFLvKCeKKJDAWCEttYwNVjB5ai6N2FxXKqYQ4,6356
@@ -248,31 +248,31 @@ quadra/schedulers/__init__.py,sha256=mQivr18c0j36hpV3Lm8nlyBVKFevWp8TtLuTfvI9kQc
 quadra/schedulers/base.py,sha256=T1EdrLOJ0i9MzWoLCkrNA0uypm7hJ-L6NFhjIXFB6NE,1462
 quadra/schedulers/warmup.py,sha256=chzzrK7OqqlicBCxiF4CqMYNrWu6nflIbRE-C86Jrw0,4962
 quadra/tasks/__init__.py,sha256=tmAfMoH0k3UC7r2pNrgbBa1Pfc3tpLl3IObFF6Z0eRE,820
-quadra/tasks/anomaly.py,sha256=rl3F39kABt5deSLOeY2s_1t2mR4x1-42VdP0flRrZMs,24583
-quadra/tasks/base.py,sha256=5Rsjdothqb4YXMIN_s98HdoPmJBfQKV0ZwnyvUuihYI,14101
-quadra/tasks/classification.py,sha256=I4NEsNVn1lOu-HLaD_hqowfnedhD4l-wZ4JPI1eD4pg,52735
-quadra/tasks/patch.py,sha256=EJvbtvlebyOtk6m47juW8XMXr1v_bLFTIKqI1KC0HRA,20244
-quadra/tasks/segmentation.py,sha256=5GF7CZjm1dQGQ-Q0kAc68GSsNv6FTxBsl_rbEZQwjmU,16213
-quadra/tasks/ssl.py,sha256=SVZeAW5xVkfeGTwMqRhGKLagMC2kHrxoOaNEDzbCZ0A,20552
+quadra/tasks/anomaly.py,sha256=RHeiM1vZF1zsva37iYdiGx_HLgdAp8lXnmUzXja69YU,24638
+quadra/tasks/base.py,sha256=piYlTFtvqH-4s4oEq4GczdAs_gL29UHAJGsOC5Sd3Bc,14187
+quadra/tasks/classification.py,sha256=05l3QM3dsU2yTWhXxNAcJ8sZM0Vbfgey-e5EV6p1TX8,52816
+quadra/tasks/patch.py,sha256=nzo8o-ei7iF1Iarvd8-c08s0Rs_lPvVPDLAbkFMx-Qw,20251
+quadra/tasks/segmentation.py,sha256=9Qy-V0Wvoofl7IrfotnSMgBIXcZd-WfZZtetyqmB0FY,16260
+quadra/tasks/ssl.py,sha256=XsaC9hbhvTA5UfHeRaaCstx9mTYacLRmgoCF5Tj9R5M,20547
 quadra/trainers/README.md,sha256=XtpbUOxwvPpOUL7E5s2JHjRgwT-CRKTxsBeUSXrg9BU,248
 quadra/trainers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 quadra/trainers/classification.py,sha256=YeJ0z7Vk0-dsMTcoKBxSdSA0rxtilEcQTp-Zq9Xi1hw,7042
 quadra/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 quadra/utils/anomaly.py,sha256=49vFvT5-4SxczsEM2Akcut_M1DDwKlOVdGv36oLTgR0,4067
-quadra/utils/classification.py,sha256=7Phbywpd1CNzwLM5F7zV9FrzdHGd6EkR6LRTRpkerTc,24882
+quadra/utils/classification.py,sha256=dKFuv4RywWhvhstOnEOnaf-6qcViUK0dTgah9m9mw2Q,24917
 quadra/utils/deprecation.py,sha256=zF_S-yqenaZxRBOudhXts0mX763WjEUWCnHd09TZnwY,852
-quadra/utils/evaluation.py,sha256=5Occ6ONsjnvMkaFFjohtgSvffxEKfjdvMbxqzgtmiHQ,18538
-quadra/utils/export.py,sha256=PoKaADeSwGPIW5TEHP3Klw_SUSV1gP1dIN4CN-AgDm0,20836
+quadra/utils/evaluation.py,sha256=oooRJPu1AaHhOwvB1Y6SFjQ645OkgrDzKtUvwWq8oq4,19005
+quadra/utils/export.py,sha256=6g--Bz1Xx8f5j_dSQTZxoiaeT_qtG-8ApfFm-CNUkmQ,20832
 quadra/utils/imaging.py,sha256=Cz7sGb_axEmnGcwQJP2djFZpIpGCPFIBGT8NWVV-OOE,866
 quadra/utils/logger.py,sha256=tQJ4xpTAFKx1g-UUm5K1x7zgoP6qoXpcUHQyu0rOr1w,556
-quadra/utils/mlflow.py,sha256=7E09JShGm2qO7bLb_8srA_RYdVAudxeDYmX7pMdjoVU,3524
+quadra/utils/mlflow.py,sha256=DVso1lxn126hil8i4tTf5WFUPJ8uJNAzNU8OXbXwOzw,3586
 quadra/utils/model_manager.py,sha256=P5JtY95p6giQ6mb4TUnWsNwUh5ClzHBillnG5SA56QY,12546
-quadra/utils/models.py,sha256=xLOT6Sorpc54dYn9XG8RpqoLupvNrMCiE_QvvIMLwHA,19756
+quadra/utils/models.py,sha256=49AXecNN7mg8uqO-YW0sLbPxbvWfTI4E4NNpTesW6HE,19699
 quadra/utils/patch/__init__.py,sha256=YenDdsI937kyAJiE0dP3_Xua8gHIoFjheoWMnpx_TGU,509
-quadra/utils/patch/dataset.py,sha256=hqM7XyPNDmI9_uJSrAxNYOasdNlEHcg-npP1S9Bb05Y,61374
-quadra/utils/patch/metrics.py,sha256=E1PeHFp10pPgkb6484fDvRLn2E9NDy9xIqEmMBeusOw,17644
+quadra/utils/patch/dataset.py,sha256=tRwrc01p0sj4nLQ-6b9mvnkTQrjtFSv5qMYiTJRSXKU,61401
+quadra/utils/patch/metrics.py,sha256=r7zxGXC2hU6EiMbfNoUmi6BC0EEUZs9Jy_mtI5Q1x5g,17693
 quadra/utils/patch/model.py,sha256=F-wbMZvM8nS_ZSYewg2SofD7H0I6DH1DBA2ACSr0fCY,5746
-quadra/utils/patch/visualization.py,sha256=HwO67vRMvhQAKN_FRFv4AAqvYMeOpg0dod3DwGpgTfo,7044
+quadra/utils/patch/visualization.py,sha256=V64SsXcQ2UhBVH2gzzrjF_OaxL58ktEo1Jdzcos3AT8,7044
 quadra/utils/resolver.py,sha256=p8t95b__htcR3hdnF9RtlWNKLTVUWYjADozYNj9lIzQ,1397
 quadra/utils/segmentation.py,sha256=rWOE1qw2RS0dpgJyHqfQURw86K6G2Hst6mpu97PI5Ac,920
 quadra/utils/tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -288,13 +288,13 @@ quadra/utils/tests/fixtures/models/classification.py,sha256=5qpyOonqK6W2LCUWEHhm
 quadra/utils/tests/fixtures/models/segmentation.py,sha256=CTNXeEPcFxFq-YcNfQi5DbbytPZwBQaZn5dQq3L41j0,765
 quadra/utils/tests/helpers.py,sha256=9PJlwozUl_lpQW-Ck-tN7sGFcgeieEd3q56aYuwMIlk,2381
 quadra/utils/tests/models.py,sha256=KbAlv_ukxaUYsyVNUO_dM0NyIosx8RpC0EVyF1HvPkM,507
-quadra/utils/utils.py,sha256=AjVsJvqACj5Bo8aOF8Itkwqhzws3bwfLnptxosbgzl4,19125
-quadra/utils/validator.py,sha256=eFCGr0ss1gYSpsL31JbsCXPZUMJAI9_H-mGodt6UGsU,4668
-quadra/utils/visualization.py,sha256=UvGHX0dumfjpT_KX3Yc1W2B5sAXXaZZWIwXQAi4sdoQ,15950
-quadra/utils/vit_explainability.py,sha256=hY0awehj6UkyBhnBlW5uWoJTsBfgow5Nll9fAqrzmMo,13337
+quadra/utils/utils.py,sha256=3tgj_tFFhKsGNJ9jrmULI9rWxFyhuUe53Y5SBJFkwSM,19124
+quadra/utils/validator.py,sha256=wmVXycB90VNyAbKBUVncFCxK4nsYiOWJIY3ISXwxYCY,4632
+quadra/utils/visualization.py,sha256=yYm7lPziUOlybxigZ2qTycNewb67Q80H4hjQGWUh788,16094
+quadra/utils/vit_explainability.py,sha256=Gh6BHaDEzWxOjJp1aqvCxLt9Rb8TXd5uKXOAx7-acUk,13351
 hydra_plugins/quadra_searchpath_plugin.py,sha256=AAn4TzR87zUK7nwSsK-KoqALiPtfQ8FvX3fgZPTGIJ0,1189
-quadra-2.2.5.dist-info/LICENSE,sha256=8cTbQtcWa02YJoSpMeV_gxj3jpMTkxvl-w3WJ5gV_QE,11342
-quadra-2.2.5.dist-info/METADATA,sha256=rG202HeIVu_pgeoNRHitD6Gs_dQj_fQcmmN1aDEUYz4,17623
-quadra-2.2.5.dist-info/WHEEL,sha256=Nq82e9rUAnEjt98J6MlVmMCZb-t9cYE2Ir1kpBmnWfs,88
-quadra-2.2.5.dist-info/entry_points.txt,sha256=sRYonBZyx-sAJeWcQNQoVQIU5lm02cnCQt6b15k0WHU,43
-quadra-2.2.5.dist-info/RECORD,,
+quadra-2.3.0.dist-info/LICENSE,sha256=8cTbQtcWa02YJoSpMeV_gxj3jpMTkxvl-w3WJ5gV_QE,11342
+quadra-2.3.0.dist-info/METADATA,sha256=0UHdlekvbPZCjDcDnFGHqKFyT8s8R4ExcaD2bDriDxQ,17598
+quadra-2.3.0.dist-info/WHEEL,sha256=Nq82e9rUAnEjt98J6MlVmMCZb-t9cYE2Ir1kpBmnWfs,88
+quadra-2.3.0.dist-info/entry_points.txt,sha256=sRYonBZyx-sAJeWcQNQoVQIU5lm02cnCQt6b15k0WHU,43
+quadra-2.3.0.dist-info/RECORD,,