vits, test=tts

paddlespeech/t2s/models/vits/vits.py

@@ -29,7 +29,7 @@ from paddlespeech.t2s.models.hifigan import HiFiGANPeriodDiscriminator
 from paddlespeech.t2s.models.hifigan import HiFiGANScaleDiscriminator
 from paddlespeech.t2s.models.vits.generator import VITSGenerator
 from paddlespeech.utils.initialize import _calculate_fan_in_and_fan_out
-from paddlespeech.utils.initialize import kaiming_normal_
+from paddlespeech.utils.initialize import kaiming_uniform_
 from paddlespeech.utils.initialize import normal_
 from paddlespeech.utils.initialize import ones_
 from paddlespeech.utils.initialize import uniform_
@@ -221,6 +221,8 @@ class VITS(nn.Layer):
         self.reuse_cache_dis = True
 
         self.reset_parameters()
+        self.generator.decoder.reset_parameters()
+        # print("VITS===============")
 
     def forward(
             self,
@@ -510,14 +512,14 @@ class VITS(nn.Layer):
         def _reset_parameters(module):
             if isinstance(module, nn.Conv1D) or isinstance(module,
                                                            nn.Conv1DTranspose):
-                kaiming_normal_(module.weight, mode="fan_out")
+                kaiming_uniform_(module.weight, a=math.sqrt(5), mode="fan_out")
                 if module.bias is not None:
                     fan_in, _ = _calculate_fan_in_and_fan_out(module.weight)
                     bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
                     uniform_(module.bias, -bound, bound)
             if isinstance(module, nn.Conv2D) or isinstance(module,
                                                            nn.Conv2DTranspose):
-                kaiming_normal_(module.weight, mode="fan_out")
+                kaiming_uniform_(module.weight, a=math.sqrt(5), mode="fan_out")
                 if module.bias is not None:
                     fan_in, _ = _calculate_fan_in_and_fan_out(module.weight)
                     bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0
@@ -527,7 +529,7 @@ class VITS(nn.Layer):
                 ones_(module.weight)
                 zeros_(module.bias)
             if isinstance(module, nn.Linear):
-                kaiming_normal_(module.weight, a=math.sqrt(5))
+                kaiming_uniform_(module.weight, a=math.sqrt(5))
                 if module.bias is not None:
                     fan_in, _ = _calculate_fan_in_and_fan_out(module.weight)
                     bound = 1 / math.sqrt(fan_in) if fan_in > 0 else 0