1

networks.py

@@ -236,11 +236,16 @@ class select_network(nn.Module):
                     padding_mask = torch.zeros_like(mixture).bool()
                     a_ref = self.BEATs_model.extract_features(mixture, padding_mask=padding_mask)[0]
                     a_ref = a_ref.transpose(1,2)
-                return self.sep_network(mixture, t_ref, a_ref.clone().detach())
+                return self.forword_step(mixture, t_ref, a_ref.clone().detach())
 
         return self.sep_network(mixture, t_ref)
 
 
+    def forword_step(self, mixture, t_ref, a_ref):
+        return self.sep_network(mixture, t_ref, a_ref)
+
+
+
 
 class network_wrapper(SpeechModel):
     def __init__(self, args):

requirements.txt

@@ -1,3 +1,4 @@
+gradio==3.44.4
 torchaudio
 torchinfo
 torchvision
@@ -5,7 +6,7 @@ librosa==0.10.2.post1
 numpy==1.26.3
 yamlargparse
 soundfile
-opencv-python
+opencv-python==4.10.0.84
 ffmpeg-python
 scikit-learn==1.5.1
 scipy