Merge branch 'dev' into merging-metatensor

Signed-off-by: Wenqi Li <[email protected]>

Author: wyli

.github/workflows/weekly-preview.yml

@@ -33,7 +33,7 @@ jobs:
         export YEAR_WEEK=$(date +'%y%U')
         echo "Year week for tag is ${YEAR_WEEK}"
         if ! [[ $YEAR_WEEK =~ ^[0-9]{4}$ ]] ; then echo "Wrong 'year week' format.  Should be 4 digits."; exit 1 ; fi
-        git tag "0.9.dev${YEAR_WEEK}"
+        git tag "0.10.dev${YEAR_WEEK}"
         git log -1
         git tag --list
         python setup.py sdist bdist_wheel

CITATION.cff

@@ -6,8 +6,8 @@ title: "MONAI: Medical Open Network for AI"
 abstract: "AI Toolkit for Healthcare Imaging"
 authors:
   - name: "MONAI Consortium"
-date-released: 2022-02-16
-version: "0.8.1"
+date-released: 2022-06-13
+version: "0.9.0"
 identifiers:
   - description: "This DOI represents all versions of MONAI, and will always resolve to the latest one."
     type: doi

Dockerfile

@@ -38,11 +38,11 @@ RUN BUILD_MONAI=1 FORCE_CUDA=1 python setup.py develop \
 
 # NGC Client
 WORKDIR /opt/tools
-ARG NGC_CLI_URI="https://ngc.nvidia.com/downloads/ngccli_cat_linux.zip"
-RUN wget -q ${NGC_CLI_URI} && \
-    unzip ngccli_cat_linux.zip && chmod u+x ngc && \
-    md5sum -c ngc.md5 && \
-    rm -rf ngccli_cat_linux.zip ngc.md5
+ARG NGC_CLI_URI="https://ngc.nvidia.com/downloads/ngccli_linux.zip"
+RUN wget -q ${NGC_CLI_URI} && unzip ngccli_linux.zip && chmod u+x ngc-cli/ngc && \
+    find ngc-cli/ -type f -exec md5sum {} + | LC_ALL=C sort | md5sum -c ngc-cli.md5 && \
+    rm -rf ngccli_linux.zip ngc-cli.md5
+ENV PATH=${PATH}:/opt/tools:/opt/tools/ngc-cli
 RUN apt-get update \
   && DEBIAN_FRONTEND="noninteractive" apt-get install -y libopenslide0  \
   && rm -rf /var/lib/apt/lists/*

docs/source/mb_specification.rst

@@ -57,7 +57,7 @@ metadata.json File
 
 This file contains the metadata information relating to the model, including what the shape and format of inputs and outputs are, what the meaning of the outputs are, what type of model is present, and other information. The JSON structure is a dictionary containing a defined set of keys with additional user-specified keys. The mandatory keys are as follows:
 
-* **version**: version of the stored model, this allows multiple versions of the same model to be differentiated.
+* **version**: version of the stored model, this allows multiple versions of the same model to be differentiated. Versions should follow semantic versioning and contain only characters valid in filenames as we may include the version to construct bundle file name.
 * **monai_version**: version of MONAI the bundle was generated on, later versions expected to work.
 * **pytorch_version**: version of Pytorch the bundle was generated on, later versions expected to work.
 * **numpy_version**: version of Numpy the bundle was generated on, later versions expected to work.

monai/metrics/generalized_dice.py

@@ -169,9 +169,15 @@ def compute_generalized_dice(
     numer = 2.0 * (intersection * w).sum(dim=1)
     denom = (denominator * w).sum(dim=1)
 
-    # Compute the score. Where the denominator (and numerator) is 0, score is 1
+    # Compute the score
     generalized_dice_score = numer / denom
-    generalized_dice_score[denom == 0] = 1
 
-    # Compute the final score, replacing nans (where denominator is 0)
+    # Handle zero deivision. Where denom == 0 and the prediction volume is 0, score is 1.
+    # Where denom == 0 but the prediction volume is not 0, score is 0
+    y_pred_o = y_pred_o.sum(dim=-1)
+    denom_zeros = denom == 0
+    generalized_dice_score[denom_zeros] = torch.where(
+        (y_pred_o == 0)[denom_zeros], torch.tensor(1.0), torch.tensor(0.0)
+    )
+
     return generalized_dice_score

monai/transforms/spatial/array.py

@@ -1345,9 +1345,9 @@ class RandRotate(RandomizableTransform, InvertibleTransform):
         range_x: Range of rotation angle in radians in the plane defined by the first and second axes.
             If single number, angle is uniformly sampled from (-range_x, range_x).
         range_y: Range of rotation angle in radians in the plane defined by the first and third axes.
-            If single number, angle is uniformly sampled from (-range_y, range_y).
+            If single number, angle is uniformly sampled from (-range_y, range_y). only work for 3D data.
         range_z: Range of rotation angle in radians in the plane defined by the second and third axes.
-            If single number, angle is uniformly sampled from (-range_z, range_z).
+            If single number, angle is uniformly sampled from (-range_z, range_z). only work for 3D data.
         prob: Probability of rotation.
         keep_size: If it is False, the output shape is adapted so that the
             input array is contained completely in the output.

monai/transforms/spatial/dictionary.py

@@ -1332,9 +1332,9 @@ class RandRotated(RandomizableTransform, MapTransform, InvertibleTransform):
         range_x: Range of rotation angle in radians in the plane defined by the first and second axes.
             If single number, angle is uniformly sampled from (-range_x, range_x).
         range_y: Range of rotation angle in radians in the plane defined by the first and third axes.
-            If single number, angle is uniformly sampled from (-range_y, range_y).
+            If single number, angle is uniformly sampled from (-range_y, range_y). only work for 3D data.
         range_z: Range of rotation angle in radians in the plane defined by the second and third axes.
-            If single number, angle is uniformly sampled from (-range_z, range_z).
+            If single number, angle is uniformly sampled from (-range_z, range_z). only work for 3D data.
         prob: Probability of rotation.
         keep_size: If it is False, the output shape is adapted so that the
             input array is contained completely in the output.

tests/test_compute_generalized_dice.py

@@ -107,12 +107,16 @@
 
 TEST_CASE_6 = [{"y": torch.ones((2, 2, 3, 3)), "y_pred": torch.ones((2, 2, 3, 3))}, [1.0000, 1.0000]]
 
-TEST_CASE_7 = [{"y": torch.zeros((2, 2, 3, 3)), "y_pred": torch.ones((2, 2, 3, 3))}, [1.0000, 1.0000]]
+TEST_CASE_7 = [{"y": torch.zeros((2, 2, 3, 3)), "y_pred": torch.ones((2, 2, 3, 3))}, [0.0000, 0.0000]]
+
+TEST_CASE_8 = [{"y": torch.ones((2, 2, 3, 3)), "y_pred": torch.zeros((2, 2, 3, 3))}, [0.0000, 0.0000]]
+
+TEST_CASE_9 = [{"y": torch.zeros((2, 2, 3, 3)), "y_pred": torch.zeros((2, 2, 3, 3))}, [1.0000, 1.0000]]
 
 
 class TestComputeMeanDice(unittest.TestCase):
     # Functional part tests
-    @parameterized.expand([TEST_CASE_1, TEST_CASE_2, TEST_CASE_6, TEST_CASE_7])
+    @parameterized.expand([TEST_CASE_1, TEST_CASE_2, TEST_CASE_6, TEST_CASE_7, TEST_CASE_8, TEST_CASE_9])
     def test_value(self, input_data, expected_value):
         result = compute_generalized_dice(**input_data)
         np.testing.assert_allclose(result.cpu().numpy(), expected_value, atol=1e-4)

tests/test_detection_coco_metrics.py

@@ -59,7 +59,7 @@ def test_coco_run(self):
             gt_classes=[val_data_i["labels"].numpy() for val_data_i in val_targets_all],
         )
         val_epoch_metric_dict = coco_metric(results_metric)[0]
-        np.testing.assert_array_less([-0.01], [sum(val_epoch_metric_dict.values())])
+        np.testing.assert_array_less([-16.01], [sum(val_epoch_metric_dict.values())])
 
 
 if __name__ == "__main__":