add attention mask

whisper.py

@@ -21,7 +21,8 @@ def generate(audio_path):
     input_speech = input_audio[0]
 
     input_features = processor(input_speech, 
-                                    sampling_rate=16_000, 
+                                    sampling_rate=16_000,
+                                    return_attention_mask=True,
                                     return_tensors="pt", torch_dtype=torch_dtype).input_features.to(device)
     
     pred_ids = model.generate(input_features,