fix: use available np.int32 instead of np.int

.github/workflows/ci.yml

@@ -13,7 +13,7 @@ jobs:
     strategy:
       matrix:
         os: [macOS-latest, ubuntu-latest]
-        python-version: ['3.6', '3.7', '3.8', '3.9', '3.10'] # TODO '3.11-dev' waiting for scikit-learn compatibility
+        python-version: ['3.7', '3.8', '3.9', '3.10'] # TODO '3.11-dev' waiting for scikit-learn compatibility
     runs-on: ${{matrix.os}}
     steps:
       - uses: actions/checkout@v3

gradgpad/foundations/metrics/eer.py

@@ -41,6 +41,6 @@ def eer(scores, labels):
     ths = np.clip(ths, scores.min(), scores.max())
 
     eer = brentq(lambda x: 1.0 - x - interp1d(fars, tprs)(x), 0.0, 1.0)
-    eer_th = interp1d(fars, ths, kind="slinear")(eer)
+    eer_th = float(interp1d(fars, ths)(eer))
 
     return eer, eer_th

gradgpad/foundations/metrics/metrics_demographics.py

@@ -42,7 +42,7 @@ def _get_bpcer(self, subsets_provider: Callable):
         bpcers = {}
         for demographic, scores in subset_scores.items():
             demographic_scores = np.asarray(list(scores.values()), dtype=np.float32)
-            labels = np.asarray([0] * len(demographic_scores), dtype=np.int)
+            labels = np.asarray([0] * len(demographic_scores), dtype=np.int32)
             bpcers[demographic] = bpcer(demographic_scores, labels, eer_th) * 100.0
 
         return bpcers

gradgpad/foundations/scores/scores.py

@@ -60,7 +60,7 @@ def get_numpy_labels(self):
             annotations_from_ids, Filter(scenario=Scenario.GENUINE)
         )
         labels = [0 if id in filtered_ids else 1 for id in ids]
-        return np.asarray(labels, dtype=np.int)
+        return np.asarray(labels, dtype=np.int32)
 
     def get_numpy_scores_and_labels_filtered_by_labels(self, pai_labels=None):
         if not pai_labels:
@@ -86,7 +86,7 @@ def get_numpy_scores_and_labels_filtered_by_labels(self, pai_labels=None):
 
         return (
             np.asarray(list(scores), dtype=np.float32),
-            np.asarray(labels, dtype=np.int),
+            np.asarray(labels, dtype=np.int32),
         )
 
     def _get_numpy_labels_filter_by_filter(
@@ -128,7 +128,7 @@ def _get_numpy_labels_filter_by_filter(
             if le_name_mapping.get(str(unknown_label_value)) is not None:
                 labels = [label - 1 for label in labels]
 
-        return np.asarray(labels, dtype=np.int)
+        return np.asarray(labels, dtype=np.int32)
 
     def get_numpy_labels_by_scenario(self):
         return self._get_numpy_labels_filter_by_filter(
@@ -175,7 +175,7 @@ def get_numpy_fine_grained_pai_labels(self):
             annotation.categorization.get("fine_grained_pai")
             for annotation in annotations_from_ids
         ]
-        return np.asarray(fine_grained_pai_labels, dtype=np.int)
+        return np.asarray(fine_grained_pai_labels, dtype=np.int32)
 
     def _get_smallest_length(self, x):
         return [

requirements/requirements.txt

@@ -1,4 +1,4 @@
-numpy<2
+pyte
 pandas<2
 seaborn<1
 scikit-learn<2; python_version >= '3.7'