correct extract embeddings

nnet/xsets.py

@@ -147,9 +147,10 @@ class SideSampler(torch.utils.data.Sampler):
             self.index_iterator[idx] = self.labels_to_indices[value][self.segment_cursors[value]]
             self.segment_cursors[value] += 1
 
-        self.index_iterator = numpy.repeat(self.index_iterator, self.num_replicas)
-        #self.index_iterator = self.index_iterator.reshape(-1, self.num_process * self.examples_per_speaker)[:, self.rank * self.examples_per_speaker:(self.rank + 1) * self.examples_per_speaker].flatten()
-        self.index_iterator = self.index_iterator.reshape(-1, self.num_process * self.examples_per_speaker * self.num_replicas)[:, self.rank * self.examples_per_speaker * self.num_replicas:(self.rank + 1) * self.examples_per_speaker * self.num_replicas].flatten()
+        print(f"num_replicas = {self.num_replicas}")
+        #self.index_iterator = numpy.repeat(self.index_iterator, self.num_replicas)
+        self.index_iterator = self.index_iterator.reshape(-1, self.num_process * self.examples_per_speaker)[:, self.rank * self.examples_per_speaker:(self.rank + 1) * self.examples_per_speaker].flatten()
+        #self.index_iterator = self.index_iterator.reshape(-1, self.num_process * self.examples_per_speaker * self.num_replicas)[:, self.rank * self.examples_per_speaker * self.num_replicas:(self.rank + 1) * self.examples_per_speaker * self.num_replicas].flatten()
 
         return iter(self.index_iterator)
 

nnet/xvector.py

@@ -1675,9 +1675,12 @@ def extract_embeddings(idmap_name,
                         _, vec = model(x=td.to(device), is_eval=True)
                         embed.append(vec.detach().cpu())
 
-                    modelset += [mod,] *  data.shape[0]
-                    segset += [seg,] *  data.shape[0]
-                    starts += [numpy.arange(start, start + vec.shape[0] * win_shift , win_shift),]
+
+                    
+                    #modelset.extend([mod,] *  data.shape[0])
+                    modelset.extend(mod *  data.shape[0])
+                    segset.extend(seg *  data.shape[0])
+                    starts.extend(numpy.arange(start, start + vec.shape[0] * win_shift , win_shift))
 
         embeddings = StatServer()
         embeddings.modelset = numpy.array(modelset).astype('>U')