Update app.py

app.py

@@ -39,8 +39,7 @@ from transformers import pipeline, AutoModelForAudioClassification, AutoProcesso
 
 
 # Set device and dtype
-#device = "cuda:0" if torch.cuda.is_available() else "cpu"
-device="cpu"
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
 torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
 lid_model_id = "facebook/mms-lid-126"
 lid_pipeline = pipeline("audio-classification", model=lid_model_id,device=device)
@@ -186,7 +185,7 @@ def generate_response(transcription):
             {"role": "user", "content": transcription},
         ]
         tokenized_chat = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
-        input_ids = tokenized_chat[0].to('cuda')
+        input_ids = tokenized_chat[0].to(device)
         if len(input_ids.shape) == 1:
             input_ids = input_ids.unsqueeze(0)
         with torch.no_grad():