Compute classification metrics locally via numpy

.pre-commit-config.yaml

@@ -29,22 +29,23 @@ repos:
         args: [--line-length=79]
 
   - repo: https://github.com/RobertCraigie/pyright-python
-    rev: v1.1.350
+    rev: v1.1.370
     hooks:
       - id: pyright
         additional_dependencies:
           [
             "requests",
             "Pillow >= 9.1.0",
             "numpy",
+            "pandas >= 2.2.2",
             "pytest",
             "python-dotenv",
             "SQLAlchemy>=2.0",
             "fastapi[all]>=0.100.0",
             "importlib_metadata; python_version < '3.8'",
             "pydantic-settings",
             "tqdm",
-            "pandas",
+            "pandas >= 2.2.2",
             "packaging",
             "PyJWT[crypto]",
             "structlog",

api/pyproject.toml

@@ -13,6 +13,7 @@ dependencies = [
     "SQLAlchemy>=2.0",
     "Pillow >= 9.1.0",
     "numpy",
+    "pandas>=2.2.2",
     "python-dotenv",
     "pydantic-settings",
     "structlog",

api/tests/functional-tests/crud/test_create_delete.py

@@ -1121,9 +1121,11 @@ def method_to_test(
             dataset_names=["test_dataset"],
             model_names=["test_model"],
             filters=schemas.Filter(
-                annotations=schemas.LogicalFunction.and_(*conditions)
-                if conditions
-                else None,
+                annotations=(
+                    schemas.LogicalFunction.and_(*conditions)
+                    if conditions
+                    else None
+                ),
                 labels=schemas.Condition(
                     lhs=schemas.Symbol(name=schemas.SupportedSymbol.LABEL_KEY),
                     rhs=schemas.Value.infer(label_key),

api/tests/unit-tests/backend/metrics/test_detection.py

@@ -17,6 +17,9 @@ def test__calculate_101_pt_interp():
     # make sure we get back 0 if we don't pass any precisions
     assert _calculate_101_pt_interp([], []) == 0
 
+    # get back -1 if all recalls and precisions are -1
+    assert _calculate_101_pt_interp([-1, -1], [-1, -1]) == -1
+
 
 def test__compute_mean_detection_metrics_from_aps():
     # make sure we get back 0 if we don't pass any precisions