Update app.py

app.py

@@ -2,7 +2,8 @@ import gradio as gr
 import spaces
 import yaml
 import torch
-import librosa
+# import librosa
+import torchaudio
 from diffusers import DDIMScheduler
 from transformers import AutoProcessor, ClapModel
 from model.udit import UDiT
@@ -98,7 +99,13 @@ def sample_diffusion(mixture, timbre, ddim_steps=50, eta=0, seed=2023, guidance_
 @spaces.GPU
 def tse(gt_file_input, text_input, num_infer_steps, eta, seed, guidance_scale, guidance_rescale):
     with torch.no_grad():
-        mixture, _ = librosa.load(gt_file_input, 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
+            
         # Check the length of the audio in samples
         current_length = len(mixture)
         target_length = sample_rate * 10