clean

features_server.py

@@ -221,7 +221,7 @@ class FeaturesServer(object):
             feat = pca_dct(feat, self.dct_pca_config[0], self.dct_pca_config[1], self.dct_pca_config[2])
         elif self.sdc:
             feat = shifted_delta_cepstral(feat, d=self.sdc_config[0], p=self.sdc_config[1], k=self.sdc_config[2])
-        
+    
         # Apply a mask on the features
         if self.mask is not None:
             feat = self._mask(feat)
@@ -488,6 +488,7 @@ class FeaturesServer(object):
             feat, label = self.post_processing(feat, label, global_mean, global_std)
         else:
             feat, label = self.post_processing(feat, label)
+
         return feat, label
 
     def get_features_per_speaker(self, show, idmap, channel=0, input_feature_filename=None, label=None):

nnet/xvector.py

@@ -1687,6 +1687,7 @@ def extract_embeddings(idmap_name,
                        model_filename,
                        data_root_name,
                        device,
+                       batch_size=1,
                        file_extension="wav",
                        transform_pipeline={},
                        sliding_window=False,
@@ -1712,6 +1713,10 @@ def extract_embeddings(idmap_name,
     :param mixed_precision:
     :return:
     """
+
+    if sliding_window:
+        batch_size = 1
+
     # Load the model
     if isinstance(model_filename, str):
         checkpoint = torch.load(model_filename, map_location=device)
@@ -1741,7 +1746,7 @@ def extract_embeddings(idmap_name,
                        )
 
     dataloader = DataLoader(dataset,
-                            batch_size=1,
+                            batch_size=batch_size,
                             shuffle=False,
                             drop_last=False,
                             pin_memory=True,