Merge branch 'main' into v2

Author: samet-akcay

.github/actions/pytest/action.yaml

@@ -180,7 +180,7 @@ runs:
 
     - name: Upload test results
       if: always() && steps.test-execution.outcome == 'failure'
-      uses: actions/upload-artifact@v3
+      uses: actions/upload-artifact@v4
       with:
         name: pytest-results-${{ inputs.test-type }}
         path: pytest_output.log

.github/workflows/pr.yaml

@@ -40,7 +40,9 @@ name: PR Checks
 
 on:
   pull_request:
-    branches: [main, "feature/**"]
+    branches: ["main", "feature/**"]
+  merge_group:
+    branches: [main]
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}

.github/workflows/pre_merge.yml

@@ -2,7 +2,7 @@ name: Pre-Merge Checks
 permissions: read-all
 
 on:
-  push:
+  merge_group:
     branches: [main]
   pull_request:
     types:

examples/notebooks/700_metrics/701e_aupimo_advanced_iv.ipynb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:adee94133a01a9b5502d2c77e0016aee4e1f2954c28eb5b7d928f5df66db42a7
-size 378257
+oid sha256:83fd473e527968ff540f506b2e63f013e4216e57f488fb1f1e6c8f8ef5632572
+size 378250

pyproject.toml

@@ -191,7 +191,6 @@ lint.ignore = [
     # End of disable rules
 
     # flake8-annotations
-    "ANN101", # Missing-type-self
     "ANN002", # Missing type annotation for *args
     "ANN003", # Missing type annotation for **kwargs
 

src/anomalib/engine/engine.py

@@ -728,6 +728,7 @@ def export(
         model: AnomalibModule,
         export_type: ExportType | str,
         export_root: str | Path | None = None,
+        model_file_name: str = "model",
         input_size: tuple[int, int] | None = None,
         compression_type: CompressionType | None = None,
         datamodule: AnomalibDataModule | None = None,
@@ -742,6 +743,8 @@ def export(
             export_type (ExportType): Export type.
             export_root (str | Path | None, optional): Path to the output directory. If it is not set, the model is
                 exported to trainer.default_root_dir. Defaults to None.
+            model_file_name (str = "model"): Name of the exported model file. If it is not set, the model is
+                is called "model". Defaults to "model".
             input_size (tuple[int, int] | None, optional): A statis input shape for the model, which is exported to ONNX
                 and OpenVINO format. Defaults to None.
             compression_type (CompressionType | None, optional): Compression type for OpenVINO exporting only.
@@ -798,15 +801,18 @@ def export(
         if export_type == ExportType.TORCH:
             exported_model_path = model.to_torch(
                 export_root=export_root,
+                model_file_name=model_file_name,
             )
         elif export_type == ExportType.ONNX:
             exported_model_path = model.to_onnx(
                 export_root=export_root,
+                model_file_name=model_file_name,
                 input_size=input_size,
             )
         elif export_type == ExportType.OPENVINO:
             exported_model_path = model.to_openvino(
                 export_root=export_root,
+                model_file_name=model_file_name,
                 input_size=input_size,
                 compression_type=compression_type,
                 datamodule=datamodule,

src/anomalib/metrics/min_max.py

@@ -85,7 +85,7 @@ def update(self, predictions: torch.Tensor, *args, **kwargs) -> None:
         del args, kwargs  # These variables are not used.
 
         self.min = torch.min(self.min, torch.min(predictions))
-        self.max = torch.max(self.min, torch.max(predictions))
+        self.max = torch.max(self.max, torch.max(predictions))
 
     def compute(self) -> tuple[torch.Tensor, torch.Tensor]:
         """Compute final minimum and maximum values.

src/anomalib/models/components/base/export_mixin.py

@@ -80,11 +80,13 @@ class ExportMixin:
     def to_torch(
         self,
         export_root: Path | str,
+        model_file_name: str = "model",
     ) -> Path:
         """Export model to PyTorch format.
 
         Args:
             export_root (Path | str): Path to the output folder
+            model_file_name (str): Name of the exported model
 
         Returns:
             Path: Path to the exported PyTorch model (.pt file)
@@ -99,7 +101,7 @@ def to_torch(
             PosixPath('./exports/weights/torch/model.pt')
         """
         export_root = _create_export_root(export_root, ExportType.TORCH)
-        pt_model_path = export_root / "model.pt"
+        pt_model_path = export_root / (model_file_name + ".pt")
         torch.save(
             obj={"model": self},
             f=pt_model_path,
@@ -109,12 +111,14 @@ def to_torch(
     def to_onnx(
         self,
         export_root: Path | str,
+        model_file_name: str = "model",
         input_size: tuple[int, int] | None = None,
     ) -> Path:
         """Export model to ONNX format.
 
         Args:
             export_root (Path | str): Path to the output folder
+            model_file_name (str): Name of the exported model.
             input_size (tuple[int, int] | None): Input image dimensions (height, width).
                 If ``None``, uses dynamic input shape. Defaults to ``None``
 
@@ -143,7 +147,7 @@ def to_onnx(
             if input_size
             else {"input": {0: "batch_size", 2: "height", 3: "weight"}, "output": {0: "batch_size"}}
         )
-        onnx_path = export_root / "model.onnx"
+        onnx_path = export_root / (model_file_name + ".onnx")
         # apply pass through the model to get the output names
         assert isinstance(self, LightningModule)  # mypy
         output_names = [name for name, value in self.eval()(input_shape)._asdict().items() if value is not None]
@@ -162,6 +166,7 @@ def to_onnx(
     def to_openvino(
         self,
         export_root: Path | str,
+        model_file_name: str = "model",
         input_size: tuple[int, int] | None = None,
         compression_type: CompressionType | None = None,
         datamodule: AnomalibDataModule | None = None,
@@ -173,6 +178,7 @@ def to_openvino(
 
         Args:
             export_root (Path | str): Path to the output folder
+            model_file_name (str): Name of the exported model
             input_size (tuple[int, int] | None): Input image dimensions (height, width).
                 If ``None``, uses dynamic input shape. Defaults to ``None``
             compression_type (CompressionType | None): Type of compression to apply.
@@ -218,9 +224,9 @@ def to_openvino(
         import openvino as ov
 
         with TemporaryDirectory() as onnx_directory:
-            model_path = self.to_onnx(onnx_directory, input_size)
+            model_path = self.to_onnx(onnx_directory, model_file_name, input_size)
             export_root = _create_export_root(export_root, ExportType.OPENVINO)
-            ov_model_path = export_root / "model.xml"
+            ov_model_path = export_root / (model_file_name + ".xml")
             ov_args = {} if ov_args is None else ov_args
 
             model = ov.convert_model(model_path, **ov_args)

tests/unit/metrics/test_min_max.py

@@ -0,0 +1,67 @@
+"""Test MinMax metric."""
+
+# Copyright (C) 2025 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import torch
+
+from anomalib.metrics import MinMax  # Assuming the metric is part of `anomalib`
+
+
+def test_initialization() -> None:
+    """Test if the metric initializes with correct default values."""
+    metric = MinMax()
+    assert torch.isinf(metric.min), "Initial min should be positive infinity."
+    assert metric.min > 0, "Initial min should be positive infinity."
+    assert torch.isinf(metric.max), "Initial max should be negative infinity."
+    assert metric.max < 0, "Initial max should be negative infinity."
+
+
+def test_update_single_batch() -> None:
+    """Test updating the metric with a single batch."""
+    metric = MinMax()
+    batch = torch.tensor([1.0, 2.0, 3.0, -1.0])
+    metric.update(batch)
+
+    assert metric.min.item() == -1.0, "Min should be -1.0 after single batch update."
+    assert metric.max.item() == 3.0, "Max should be 3.0 after single batch update."
+
+
+def test_update_multiple_batches() -> None:
+    """Test updating the metric with multiple batches."""
+    metric = MinMax()
+    batch1 = torch.tensor([0.5, 1.5, 3.0])
+    batch2 = torch.tensor([-0.5, 0.0, 2.5])
+
+    metric.update(batch1)
+    metric.update(batch2)
+
+    assert metric.min.item() == -0.5, "Min should be -0.5 after multiple batch updates."
+    assert metric.max.item() == 3.0, "Max should be 3.0 after multiple batch updates."
+
+
+def test_compute() -> None:
+    """Test computation of the min and max values after updates."""
+    metric = MinMax()
+    batch1 = torch.tensor([1.0, 2.0])
+    batch2 = torch.tensor([-1.0, 0.0])
+
+    metric.update(batch1)
+    metric.update(batch2)
+
+    min_val, max_val = metric.compute()
+
+    assert min_val.item() == -1.0, "Computed min should be -1.0."
+    assert max_val.item() == 2.0, "Computed max should be 2.0."
+
+
+def test_no_updates() -> None:
+    """Test behavior when no updates are made to the metric."""
+    metric = MinMax()
+
+    min_val, max_val = metric.compute()
+
+    assert torch.isinf(min_val), "Min should remain positive infinity with no updates."
+    assert min_val > 0, "Min should remain positive infinity with no updates."
+    assert torch.isinf(max_val), "Max should remain negative infinity with no updates."
+    assert max_val < 0, "Max should remain negative infinity with no updates."