fix: fix the bug which leads qwen2_5_vl to crash with mixed_chunk (#11330)

Signed-off-by: PanJason <pyyjason@gmail.com>
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
Co-authored-by: Yuan Luo <yuan.luo@hotmail.com>

python/sglang/srt/model_executor/forward_batch_info.py

@@ -557,6 +557,25 @@ class ForwardBatch:
 
         self.mrope_positions = next_input_positions
 
+    def _expand_mrope_from_input(
+        self,
+        mm_input: MultimodalInputs,
+        seq_len: int,
+        device: torch.device,
+    ) -> torch.Tensor:
+        if mm_input.mrope_position_delta.device.type != device:
+            # transfer mrope_position_delta to device when the first running,
+            # avoiding successvie host-to-device data transfer
+            mm_input.mrope_position_delta = mm_input.mrope_position_delta.to(
+                device, non_blocking=True
+            )
+
+        mrope_position_deltas = mm_input.mrope_position_delta.flatten()
+        mrope_positions = (
+            (mrope_position_deltas + seq_len - 1).unsqueeze(0).repeat(3, 1)
+        )
+        return mrope_positions
+
     def _compute_mrope_positions(
         self, model_runner: ModelRunner, batch: ModelWorkerBatch
     ):
@@ -575,20 +594,10 @@ class ForwardBatch:
                         device=model_runner.device,
                     )
                 else:
-                    if mm_input.mrope_position_delta.device.type != model_runner.device:
-                        # transfer mrope_position_delta to device when the first running,
-                        # avoiding successvie host-to-device data transfer
-                        mm_input.mrope_position_delta = (
-                            mm_input.mrope_position_delta.to(
-                                model_runner.device, non_blocking=True
-                            )
-                        )
-                    mrope_position_deltas = mm_input.mrope_position_delta.flatten()
-                    mrope_positions_list[batch_idx] = (
-                        (mrope_position_deltas + self.seq_lens[batch_idx] - 1)
-                        .unsqueeze(0)
-                        .repeat(3, 1)
+                    mrope_positions = self._expand_mrope_from_input(
+                        mm_input, self.seq_lens[batch_idx], model_runner.device
                     )
+                    mrope_positions_list[batch_idx] = mrope_positions
             elif self.forward_mode.is_extend():
                 extend_seq_len, extend_prefix_len = (
                     batch.extend_seq_lens[batch_idx],
@@ -613,6 +622,10 @@ class ForwardBatch:
                         :,
                         extend_prefix_len : extend_prefix_len + extend_seq_len,
                     ]
+                    if mrope_positions.numel() == 0:
+                        mrope_positions = self._expand_mrope_from_input(
+                            mm_input, self.seq_lens[batch_idx], model_runner.device
+                        )
                 mrope_positions_list[batch_idx] = mrope_positions
 
         self.mrope_positions = torch.cat(