[Bugfix][CPU] fix missing input intermediate_tensors in the cpu_model_runner (#8733)

3e83c12b · Li, Jiang · GitHub · e551ca15 · 3e83c12b
Unverified Commit 3e83c12b authored Sep 23, 2024 by Li, Jiang Committed by GitHub Sep 23, 2024
Hide whitespace changes
Inline Side-by-side

Showing with 2 additions and 0 deletions

vllm/worker/cpu_model_runner.py vllm/worker/cpu_model_runner.py +2 -0

No files found.
--- a/vllm/worker/cpu_model_runner.py
+++ b/vllm/worker/cpu_model_runner.py
@@ -456,6 +456,8 @@ class CPUModelRunner(ModelRunnerBase[ModelInputForCPU]):
            model_input.attn_metadata,
            **MultiModalInputs.as_kwargs(model_input.multi_modal_kwargs or {},
                                         device=self.device),
+            "intermediate_tensors":
+            intermediate_tensors,
        }

        hidden_states = model_executable(**execute_model_kwargs)