fix

app.py

@@ -15,12 +15,14 @@ from transformers import Blip2Processor, Blip2ForConditionalGeneration
 from diffusers import DDIMScheduler
 import spaces
 
-device = 'cuda' if torch.cuda.is_available() else 'cpu'
-if device == 'cpu':
-    torch_dtype = torch.float32
-else:
-    torch_dtype = torch.float16
-
+# device = 'cuda' if torch.cuda.is_available() else 'cpu'
+# if device == 'cpu':
+#     torch_dtype = torch.float32
+# else:
+#     torch_dtype = torch.float16
+
+device = 'cuda'
+torch_dtype = torch.float16
 def memory_efficient(model):
     try:
         model.to(device)
@@ -381,6 +383,7 @@ description_md = """
 ### We introduce `Visual Style Prompting`, which reflects the style of a reference image to the images generated by a pretrained text-to-image diffusion model without finetuning or optimization (e.g., Figure N).
 ### 📖 [[Paper](https://arxiv.org/abs/2402.12974)] | ✨ [[Project page](https://curryjung.github.io/VisualStylePrompt)] | ✨ [[Code](https://github.com/naver-ai/Visual-Style-Prompting)]
 ---
+### 👉 To better reflect the style of a user's image, the higher the resolution, the better.
 ### 🔥 To try out our vanilla demo,
 1. Choose a `style reference` from the collection of images below.
 2. Enter the `text prompt`.