Update app.py

app.py

@@ -83,7 +83,7 @@ JPEG_QUALITY = 80
 GPU_MEMORY_FRACTION = 0.8
 
 # --- 1. Device ---
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+device = torch.device("cpu") #torch.device("cuda" if torch.cuda.is_available() else "cpu")
 torch.backends.cudnn.benchmark = True
 if device.type == 'cuda':
     torch.cuda.set_per_process_memory_fraction(GPU_MEMORY_FRACTION, device=0)
@@ -180,6 +180,7 @@ def handle_file(file, prompt, extra_prompt, max_new_tokens, progress=gr.Progress
 # --- 3. Inference Function ---
 @spaces.GPU
 def run_inference(img: Image.Image, prompt: str = "", max_new_tokens: int = 512) -> str:
+     model.to("cuda")
     if img.mode != "RGB":
         img = img.convert("RGB")
     prompt_text = prompt.strip()