From 6925cdbeea0d2061e47403cce10f015d99f00ab1 Mon Sep 17 00:00:00 2001
From: Isotr0py <2037008807@qq.com>
Date: Thu, 22 Aug 2024 00:23:03 +0800
Subject: [PATCH] [Bugfix][Hardware][CPU] Fix `mm_limits` initialization for
 CPU backend (#7735)

---
 vllm/worker/cpu_model_runner.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/vllm/worker/cpu_model_runner.py b/vllm/worker/cpu_model_runner.py
index 82e69b569d90d..f69afa4c43149 100644
--- a/vllm/worker/cpu_model_runner.py
+++ b/vllm/worker/cpu_model_runner.py
@@ -113,8 +113,10 @@ def __init__(
         )
 
         # Multi-modal data support
-        self.multi_modal_input_mapper = MULTIMODAL_REGISTRY \
+        self.mm_registry = MULTIMODAL_REGISTRY
+        self.multi_modal_input_mapper = self.mm_registry \
             .create_input_mapper(self.model_config)
+        self.mm_registry.init_mm_limits_per_prompt(self.model_config)
 
         # Lazy initialization.
         self.model: nn.Module  # Set after init_Model