update

examples/clean_unet_aishell/step_2_train_model.py

@@ -319,8 +319,12 @@ def main():
                 enhanced_audios_list_r = list(enhanced_audios.detach().cpu().numpy())
                 clean_audios_list_r = list(clean_audios.detach().cpu().numpy())
                 pesq_metric = batch_pesq(enhanced_audios_list_r, clean_audios_list_r)
+                if pesq_metric is None:
+                    pesq_metric = 0
+                else:
+                    pesq_metric = torch.mean(pesq_metric).item()
 
-                total_pesq_metric += pesq_metric.item()
+                total_pesq_metric += pesq_metric
                 total_loss += loss.item()
                 total_ae_loss += ae_loss.item()
                 total_sc_loss += sc_loss.item()

toolbox/torchaudio/models/clean_unet/modeling_clean_unet.py

@@ -144,7 +144,7 @@ class CleanUNet(nn.Module):
                     nn.Conv1d(channels_h, channels_h * 2, 1),
                     nn.GLU(dim=1),
                     nn.ConvTranspose1d(channels_h, channels_output, kernel_size, stride),
-                    # nn.ReLU(inplace=False)
+                    nn.ReLU(inplace=False)
                 ))
             channels_output = channels_h