Update app.py

app.py

@@ -16,7 +16,7 @@ from transformers import TextIteratorStreamer
 from threading import Thread
 import torch
 
-device = 'auto'
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
 model_id = "google/gemma-3-4b-it"
 processor = AutoProcessor.from_pretrained(model_id, use_fast=True, padding_side="left")
 model = Gemma3ForConditionalGeneration.from_pretrained(