diff --git a/vllm_omni/worker/gpu_model_runner.py b/vllm_omni/worker/gpu_model_runner.py
index efad5f3d4..893740b75 100644
--- a/vllm_omni/worker/gpu_model_runner.py
+++ b/vllm_omni/worker/gpu_model_runner.py
@@ -1120,7 +1120,7 @@ class OmniGPUModelRunner(GPUModelRunner):
                     dtype=self.dtype, device=self.device, non_blocking=True
                 )
                 start_offset = int(self.query_start_loc.cpu[req_index])
-                self.inputs_embeds[start_offset : start_offset + overlay_len].copy_(src)
+                self.inputs_embeds.gpu[start_offset : start_offset + overlay_len].copy_(src)
 
     def _update_additional_information(self, scheduler_output: "SchedulerOutput") -> None:
         for new_req in scheduler_output.scheduled_new_reqs: