Update modeling_diva.py

modeling_diva.py

@@ -252,7 +252,7 @@ class DiVAModel(PreTrainedModel):
         self, audio, text_prompt, do_sample=False, logits_processor=None, max_new_tokens=128
     ):
         inputs = self.processor(audio, return_tensors="pt", sampling_rate=16_000)
-        input_features = inputs.input_features
+        input_features = inputs.input_features.to(self.whisper_encoder.device)
         hidden_states = self.whisper_encoder(input_features=input_features)[
             "last_hidden_state"
         ]