not apply auto batch for CPU plugin

Signed-off-by: HU Yuan2 <[email protected]>

Author: tiger100256-hu

src/inference/src/dev/core_impl.cpp

@@ -1094,8 +1094,11 @@ std::shared_ptr<const ov::Model> ov::CoreImpl::apply_auto_batching(const std::sh
                            .get_property(ov::supported_properties.name(), parsed._config)
                            .as<std::vector<ov::PropertyName>>();
         auto it = std::find(metrics.begin(), metrics.end(), ov::optimal_batch_size.name());
-        if (metrics.end() == it)
+        if (metrics.end() == it) {
             return model;
+        } else if (parsed._deviceName == "CPU") {//not apply auto batch for CPU plugin
+            return model;
+        }
 
         // if applicable, the Auto-Batching is implicitly enabled via the performance hints
         bool bTputInPlg = get_plugin(parsed._deviceName)