diff --git a/llama.cpp b/llama.cpp
index 2105e3279..ce33f7fbe 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -3122,9 +3122,9 @@ size_t llama_copy_state_data(struct llama_context * ctx, uint8_t * dst) {
         if (kv_size) {
             const size_t elt_size = ggml_element_size(kv_self.k);
 
-            char buffer[4096];
+            //char buffer[4096];
 
-            ggml_context * cpy_ctx = ggml_init({ sizeof(buffer), buffer, /* no_alloc */ true });
+            ggml_context * cpy_ctx = ggml_init({ 4096, NULL, /* no_alloc */ true });
             ggml_cgraph gf{};
             gf.n_threads = 1;