fix

app.py

@@ -9,7 +9,7 @@ def generate_text(self, image: str, prompt: str) -> str:
     image = image.convert('RGB')
     image = to_tensor(image).unsqueeze(0).to(self.device)
     
-    image_code = self.visual_tokenizer.encoder(image)['bottleneck_rep']
+    image_code = self.visual_tokenizer.encoder(image.to(self.config.dtype))['bottleneck_rep']
     image_text = "".join([f"<I{x}>" for x in image_code[0].cpu().tolist()])
     
     messages = [