misc cleaning

Author: NDevanathan

.github/workflows/pre-commit.yml

@@ -16,10 +16,12 @@ jobs:
     - name: Cache uv dependencies
       uses: actions/cache@v4
       with:
-        path: .venv
-        key: ${{ runner.os }}-uv-${{ hashFiles('uv.lock') }}
+        path: |
+          .venv
+          ~/.cache/uv
+        key: ${{ runner.os }}-uv-${{ hashFiles('uv.lock', 'pyproject.toml') }}
         restore-keys: |
-          ${{ runner.os }}-uv-
+          ${{ runner.os }}-uv--
     - name: Install dependencies
       run: uv sync
     - uses: pre-commit/[email protected]

.github/workflows/release.yml

@@ -16,7 +16,9 @@ jobs:
       - name: Cache uv dependencies
         uses: actions/cache@v4
         with:
-          path: .venv
+          path: |
+            .venv
+            ~/.cache/uv
           key: ${{ runner.os }}-uv-${{ hashFiles('uv.lock', 'pyproject.toml') }}
           restore-keys: |
             ${{ runner.os }}-uv-

ls_spa/__init__.py

@@ -2,6 +2,7 @@
 
 from .ls_spa import (
     ShapleyResults,
+    SizeIncompatible,
     SizeIncompatibleError,
     error_estimates,
     ls_spa,
@@ -13,6 +14,7 @@
 
 __all__ = [
     "ShapleyResults",
+    "SizeIncompatible",
     "SizeIncompatibleError",
     "error_estimates",
     "ls_spa",

ls_spa/ls_spa.py

@@ -29,7 +29,7 @@
 from numpy import random
 
 # The maximum number of features for which we can display the full attribution.
-CAN_DISPLAY_FULL_ATTR = 5
+MAX_ATTR_DISP = 5
 
 # The maximum number of features for which we can feasibly compute the exact Shapley values.
 MAX_FEAS_EXACT_FEATS = 9
@@ -52,15 +52,19 @@ def __repr__(self) -> str:
         attr_str = ""
         coefs_str = ""
 
-        if len(self.attribution) <= CAN_DISPLAY_FULL_ATTR:
+        if len(self.attribution) <= MAX_ATTR_DISP:
             attr_str = "(" + "".join(f"{a:.2f}, " for a in self.attribution.flatten())[:-2] + ")"
             coefs_str = "(" + "".join(f"{c:.2f}, " for c in self.theta.flatten())[:-2] + ")"
         else:
             attr_str = (
-                "(" + "".join(f"{a:.2f}, " for a in self.attribution.flatten()[:5])[:-2] + ", ...)"
+                "("
+                + "".join(f"{a:.2f}, " for a in self.attribution.flatten()[:MAX_ATTR_DISP])[:-2]
+                + ", ...)"
             )
             coefs_str = (
-                "(" + "".join(f"{c:.2f}, " for c in self.theta.flatten()[:5])[:-2] + ", ...)"
+                "("
+                + "".join(f"{c:.2f}, " for c in self.theta.flatten()[:MAX_ATTR_DISP])[:-2]
+                + ", ...)"
             )
 
         return f"""
@@ -83,6 +87,11 @@ def __init__(self, message: str) -> None:
         super().__init__(self.message)
 
 
+# TODO(ndevanathan): remove this in the next major update
+# This is here for backwards compatibility
+SizeIncompatible = SizeIncompatibleError
+
+
 def validate_data(
     X_train: np.ndarray,
     X_test: np.ndarray,
@@ -410,7 +419,7 @@ def error_estimates(rng: random.Generator, cov: np.ndarray) -> tuple[np.ndarray,
     p = cov.shape[0]
     try:
         sample_diffs = rng.multivariate_normal(np.zeros(p), cov, size=2**10, method="cholesky")
-    except:  # noqa: E722
+    except (np.linalg.LinAlgError, ValueError):
         sample_diffs = rng.multivariate_normal(np.zeros(p), cov, size=2**10, method="svd")
     abs_diffs = np.abs(sample_diffs)
     norms = np.linalg.norm(sample_diffs, axis=1)

pyproject.toml

@@ -77,7 +77,6 @@ extend-exclude = [
     "notebooks",
     "paper",
     "data",
-    ".venv",
     "__marimo__",
     "__pycache__",
     "build",