Update app.py

app.py

@@ -119,12 +119,16 @@ def sample_diffusion(mixture, timbre, ddim_steps=50, eta=0, seed=2023, guidance_
 def tse(gt_file_input, text_input, num_infer_steps, eta, seed, guidance_scale, guidance_rescale):
     reset_scheduler_dtype()
     with torch.no_grad():
-        mixture, _ = librosa.load(gt_file_input, sr=sample_rate)
-        # mixture, sr = torchaudio.load(gt_file_input)
-        # if sr != sample_rate:
-        #     resampler = torchaudio.transforms.Resample(orig_freq=sr, new_freq=sample_rate)
-        #     mixture = resampler(mixture)
-        #     sr = sample_rate
+        # mixture, _ = librosa.load(gt_file_input, sr=sample_rate)
+        mixture, sr = torchaudio.load(gt_file_input)
+        if sr != sample_rate:
+            resampler = torchaudio.transforms.Resample(orig_freq=sr, new_freq=sample_rate)
+            mixture = resampler(mixture)
+            sr = sample_rate
+        if mixture.shape[0] > 1:
+            mixture = torch.mean(mixture, dim=0)
+        else:
+            mixture = mixture[0]
             
         # Check the length of the audio in samples
         current_length = len(mixture)