From 2f8cd979ecd1fa582852e7136e92ff8990b98fd8 Mon Sep 17 00:00:00 2001
From: Aaron Miller <apage43@ninjawhale.com>
Date: Sat, 1 Jul 2023 11:14:59 -0700
Subject: metal : release buffers when freeing metal context (#2062)

---
 llama.cpp | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

(limited to 'llama.cpp')

diff --git a/llama.cpp b/llama.cpp
index 69c2ab0..561accf 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -253,7 +253,13 @@ struct llama_model {
 
 struct llama_context {
     llama_context(const llama_model & model, const llama_vocab & vocab) : model(model), vocab(vocab), t_load_us(model.t_load_us), t_start_us(model.t_start_us) {}
-
+#ifdef GGML_USE_METAL
+    ~llama_context() {
+        if (ctx_metal) {
+            ggml_metal_free(ctx_metal);
+        }
+    }
+#endif
     std::mt19937 rng;
 
     bool has_evaluated_once = false;
-- 
cgit v1.2.3