update

toolbox/torchaudio/models/mpnet/discriminator.py

@@ -15,22 +15,22 @@ from toolbox.torchaudio.models.mpnet.configuration_mpnet import MPNetConfig
 from toolbox.torchaudio.models.mpnet.utils import LearnableSigmoid1d
 
 
-def cal_pesq(clean, noisy, sr=16000):
-    try:
-        pesq_score = pesq(sr, clean, noisy, 'wb')
-    except:
-        # error can happen due to silent period
-        pesq_score = -1
-    return pesq_score
-
-
-def batch_pesq(clean, noisy):
-    pesq_score = Parallel(n_jobs=15)(delayed(cal_pesq)(c, n) for c, n in zip(clean, noisy))
-    pesq_score = np.array(pesq_score)
-    if -1 in pesq_score:
-        return None
-    pesq_score = (pesq_score - 1) / 3.5
-    return torch.FloatTensor(pesq_score)
+# def cal_pesq(clean, noisy, sr=16000):
+#     try:
+#         pesq_score = pesq(sr, clean, noisy, 'wb')
+#     except:
+#         # error can happen due to silent period
+#         pesq_score = -1
+#     return pesq_score
+
+
+# def batch_pesq(clean, noisy):
+#     pesq_score = Parallel(n_jobs=15)(delayed(cal_pesq)(c, n) for c, n in zip(clean, noisy))
+#     pesq_score = np.array(pesq_score)
+#     if -1 in pesq_score:
+#         return None
+#     pesq_score = (pesq_score - 1) / 3.5
+#     return torch.FloatTensor(pesq_score)
 
 
 def metric_loss(metric_ref, metrics_gen):

toolbox/torchaudio/models/mpnet/modeling_mpnet.py

@@ -250,26 +250,26 @@ def anti_wrapping_function(x):
     return torch.abs(x - torch.round(x / (2 * np.pi)) * 2 * np.pi)
 
 
-def pesq_score(utts_r, utts_g, h):
-
-    pesq_score = Parallel(n_jobs=30)(delayed(eval_pesq)(
-                            utts_r[i].squeeze().cpu().numpy(),
-                            utts_g[i].squeeze().cpu().numpy(),
-                            h.sample_rate, )
-                          for i in range(len(utts_r)))
-    pesq_score = np.mean(pesq_score)
-
-    return pesq_score
-
-
-def eval_pesq(clean_utt, esti_utt, sr):
-    try:
-        mode = "nb" if sr == 8000 else "wb"
-        pesq_score = pesq(sr, clean_utt, esti_utt, mode=mode)
-    except:
-        pesq_score = -1
-
-    return pesq_score
+# def pesq_score(utts_r, utts_g, h):
+#
+#     pesq_score = Parallel(n_jobs=30)(delayed(eval_pesq)(
+#                             utts_r[i].squeeze().cpu().numpy(),
+#                             utts_g[i].squeeze().cpu().numpy(),
+#                             h.sample_rate, )
+#                           for i in range(len(utts_r)))
+#     pesq_score = np.mean(pesq_score)
+#
+#     return pesq_score
+#
+#
+# def eval_pesq(clean_utt, esti_utt, sr):
+#     try:
+#         mode = "nb" if sr == 8000 else "wb"
+#         pesq_score = pesq(sr, clean_utt, esti_utt, mode=mode)
+#     except:
+#         pesq_score = -1
+#
+#     return pesq_score
 
 
 def main():