Upload 3 files

app.py

@@ -16,8 +16,9 @@ processor = AutoProcessor.from_pretrained(MODEL_ID)
 # Load Model
 model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID, 
-    torch_dtype=torch.float16
-).to(device).eval()
+    torch_dtype=torch.bfloat16,
+    low_cpu_mem_usage=True,
+).eval()
 print("Model loaded.")
 
 @spaces.GPU