Update inference.py

inference.py

@@ -19,10 +19,11 @@ class PaliGemmaModel:
     def infer(self, image: PIL.Image.Image, text: str, max_new_tokens: int) -> str:
         inputs = self.processor(text=text, images=image, return_tensors="pt").to(self.device)
         with torch.inference_mode():
+            inputs = {k: v.to(self.device) for k, v in inputs.items()}
             generated_ids = self.model.generate(
                 **inputs,
                 max_new_tokens=max_new_tokens,
                 do_sample=False
             )
         result = self.processor.batch_decode(generated_ids, skip_special_tokens=True)
-        return result[0][len(text):].lstrip("\n")
+        return result[0][len(text):].lstrip("\n")