Add missing mean_stat_per_model method to StatObject_SB (#3029)

Mr-Neutr0n · web-flow · commit a601051f44f6 · 2026-02-10T12:52:48.000-05:00
diff --git a/speechbrain/processing/PLDA_LDA.py b/speechbrain/processing/PLDA_LDA.py
@@ -223,6 +223,19 @@ def sum_stat_per_model(self):
             session_per_model[idx] += self.get_model_stat1(model).shape[0]
         return sts_per_model, session_per_model
 
+    def mean_stat_per_model(self):
+        """Average the zero- and first-order statistics per model and store
+        them in a new StatObject_SB.
+
+        Returns
+        -------
+        a StatObject_SB object with the statistics averaged per model.
+        """
+        sts_per_model, session_per_model = self.sum_stat_per_model()
+        sts_per_model.stat0 = sts_per_model.stat0 / session_per_model[:, None]
+        sts_per_model.stat1 = sts_per_model.stat1 / session_per_model[:, None]
+        return sts_per_model
+
     def center_stat1(self, mu):
         """Center first order statistics.
 
@@ -721,10 +734,13 @@ def fast_PLDA_scoring(
     enroll_ctr = copy.deepcopy(enroll)
     test_ctr = copy.deepcopy(test)
 
-    # If models are not unique, compute the mean per model, display a warning
+    # If models are not unique, require the user to average them first
     if not numpy.unique(enroll_ctr.modelset).shape == enroll_ctr.modelset.shape:
-        # logging.warning("Enrollment models are not unique, average i-vectors")
-        enroll_ctr = enroll_ctr.mean_stat_per_model()
+        raise ValueError(
+            "Enrollment models are not unique. Call "
+            "enroll.mean_stat_per_model() before passing to "
+            "fast_PLDA_scoring() to average statistics per model."
+        )
 
     # Remove missing models and test segments
     if check_missing:
@@ -736,11 +752,6 @@ def fast_PLDA_scoring(
     enroll_ctr.center_stat1(mu)
     test_ctr.center_stat1(mu)
 
-    # If models are not unique, compute the mean per model, display a warning
-    if not numpy.unique(enroll_ctr.modelset).shape == enroll_ctr.modelset.shape:
-        # logging.warning("Enrollment models are not unique, average i-vectors")
-        enroll_ctr = enroll_ctr.mean_stat_per_model()
-
     # Compute constant component of the PLDA distribution
     invSigma = linalg.inv(Sigma)
     I_spk = numpy.eye(F.shape[1], dtype="float")
