Fix processing of stereo clips

app.py

@@ -42,14 +42,14 @@ def load_models():
 def audio_classification(audio_file):
     audio, sr = torchaudio.load(audio_file)
     audio = torchaudio.transforms.Resample(sr, cfg.sample_rate)(audio)
-    # Add dimension for batch
-    audio = audio.unsqueeze(0)
     # Convert to mono
     audio = audio.mean(0, keepdim=True)
+    # Add dimension for batch
+    audio = audio.unsqueeze(0)
     audio = audio.to(device)
 
     with torch.no_grad():
-        # Classifiy
+        # Classify
         print("Detecting effects")
         labels = torch.tensor(classifier(audio))
         labels_dict = {
@@ -62,10 +62,10 @@ def audio_classification(audio_file):
 def audio_removal(audio_file, labels, threshold):
     audio, sr = torchaudio.load(audio_file)
     audio = torchaudio.transforms.Resample(sr, cfg.sample_rate)(audio)
-    # Add dimension for batch
-    audio = audio.unsqueeze(0)
     # Convert to mono
     audio = audio.mean(0, keepdim=True)
+    # Add dimension for batch
+    audio = audio.unsqueeze(0)
     audio = audio.to(device)
 
     label_names = [f"RandomPedalboard{lab['label']}" for lab in labels["confidences"]]