From c05883f3a9370448722f6ba85e1c7d1f069f1a83 Mon Sep 17 00:00:00 2001
From: 0cc4m <picard12@live.de>
Date: Sat, 16 Dec 2023 14:49:03 +0100
Subject: [PATCH] Free model gpu buffers on exit

---
 llama.cpp | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/llama.cpp b/llama.cpp
index 73f987188..a0fc74e42 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -1451,6 +1451,9 @@ struct llama_model {
             ggml_cl_free_data(tensors_by_name[i].second);
         }
 #elif defined(GGML_USE_VULKAN)
+        for (size_t i = 0; i < tensors_by_name.size(); ++i) {
+            ggml_vk_free_data(tensors_by_name[i].second);
+        }
         ggml_vk_cleanup();
 #endif
     }