add audios files

app.py

@@ -70,11 +70,19 @@ def enhance_speech(audio_array, sampling_rate, solver, nfe, tau, denoise_before_
 
     @spaces.GPU(duration=360)
     def denoise_audio():
-        return denoise(audio_array, sampling_rate, device)
+        try:
+            return denoise(audio_array, sampling_rate, device)
+        except Exception as e:
+            print("> Error while denoising : ", str(e))
+            return audio_array, sampling_rate
 
     @spaces.GPU(duration=360)
     def enhance_audio():
-        return enhance(audio_array, sampling_rate, device, nfe=nfe, solver=solver, lambd=lambd, tau=tau)
+        try:
+            return enhance(audio_array, sampling_rate, device, nfe=nfe, solver=solver, lambd=lambd, tau=tau)
+        except Exception as e:
+            print("> Error while enhancement : ", str(e))
+            return audio_array, sampling_rate
 
     with concurrent.futures.ThreadPoolExecutor() as executor:
         future_denoise = executor.submit(denoise_audio)
@@ -119,7 +127,7 @@ def _fn(
     )
 
     # Return all outputs
-    return bambara_text, (sampling_rate, audio_array.cpu().numpy()), denoised_audio, enhanced_audio
+    return bambara_text, (sampling_rate, audio_array.numpy()), denoised_audio, enhanced_audio
 
 
 def main():

tts.py

@@ -373,7 +373,7 @@ class BambaraTTS:
         )
         end_time = time.time()
 
-        audio = torch.tensor(out["wav"]).unsqueeze(0)
+        audio = torch.tensor(out["wav"]).unsqueeze(0).cpu()
         sampling_rate = self.config.model_args.output_sample_rate
 
         self.log(f"Speech generated in {end_time - start_time:.2f} seconds.")