From d4f7df1c360d22d63f7bdabfd64a6c2409fad522 Mon Sep 17 00:00:00 2001 From: "jiang1.li" Date: Mon, 23 Sep 2024 09:37:26 +0000 Subject: [PATCH 1/2] fix cpu model runner input --- vllm/worker/cpu_model_runner.py | 1 + 1 file changed, 1 insertion(+) diff --git a/vllm/worker/cpu_model_runner.py b/vllm/worker/cpu_model_runner.py index 7b2caf497358..42da426ce17c 100644 --- a/vllm/worker/cpu_model_runner.py +++ b/vllm/worker/cpu_model_runner.py @@ -372,6 +372,7 @@ def execute_model( 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) From dc03c30cf1673a551b3606e4be9f6aa004970e57 Mon Sep 17 00:00:00 2001 From: "jiang1.li" Date: Mon, 23 Sep 2024 09:56:16 +0000 Subject: [PATCH 2/2] format --- vllm/worker/cpu_model_runner.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/vllm/worker/cpu_model_runner.py b/vllm/worker/cpu_model_runner.py index 42da426ce17c..372ef24410bb 100644 --- a/vllm/worker/cpu_model_runner.py +++ b/vllm/worker/cpu_model_runner.py @@ -372,7 +372,8 @@ def execute_model( model_input.attn_metadata, **MultiModalInputs.as_kwargs(model_input.multi_modal_kwargs or {}, device=self.device), - "intermediate_tensors": intermediate_tensors, + "intermediate_tensors": + intermediate_tensors, } hidden_states = model_executable(**execute_model_kwargs)