clean up model_runner.py

vllm/worker/model_runner.py

@@ -1218,8 +1218,7 @@ def capture_model(self, kv_caches: List[List[torch.Tensor]]) -> None:
         # Prepare dummy inputs. These will be reused for all batch sizes.
         max_batch_size = max(_BATCH_SIZES_TO_CAPTURE)
         input_tokens = torch.zeros(max_batch_size, dtype=torch.long).cuda()
-        input_positions = torch.zeros(max_batch_size, dtype=torch.long).cuda()
-
+        input_positions = torch.zeros(max_batch_size, dtype=torch.long).cuda()        
         # Prepare dummy previous_hidden_states only if needed by the model.
         # This is used by draft models such as EAGLE.
         previous_hidden_states = None