From 82e4187f95f969b90b9e27c660cb5168c3e33ec7 Mon Sep 17 00:00:00 2001
From: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
Date: Fri, 19 Apr 2024 13:16:42 +0200
Subject: [PATCH] llama: add llama_model_kv_override_free

---
 common/common.cpp |  1 -
 llama.cpp         | 18 ++++++++++++++++++
 llama.h           |  3 +++
 3 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/common/common.cpp b/common/common.cpp
index 29fa5cdbc..9838a538f 100644
--- a/common/common.cpp
+++ b/common/common.cpp
@@ -263,7 +263,6 @@ bool parse_kv_override(const char * data, std::vector<llama_model_kv_override> &
             fprintf(stderr, "%s: invalid boolean value for KV override '%s'\n", __func__, data);
             return false;
         }
-
     } else if (strncmp(sep, "str:", 4) == 0) {
         sep += 4;
         kvo.tag = LLAMA_KV_OVERRIDE_TYPE_STR;
diff --git a/llama.cpp b/llama.cpp
index 8bf1fdbc6..2907283f4 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -2929,6 +2929,16 @@ namespace GGUFMeta {
             return false;
         }
 
+        template<typename OT>
+        static typename std::enable_if<std::is_same<OT, char *>::value, char *>::type
+        try_override(T & target, const struct llama_model_kv_override * ovrd) {
+            if (validate_override(LLAMA_KV_OVERRIDE_TYPE_STR, ovrd)) {
+                target = ovrd->str_value;
+                return true;
+            }
+            return false;
+        }
+
         template<typename OT>
         static typename std::enable_if<std::is_same<OT, std::string>::value, bool>::type
         try_override(T & target, const struct llama_model_kv_override * ovrd) {
@@ -14977,6 +14987,14 @@ void llama_free_model(struct llama_model * model) {
     delete model;
 }
 
+void llama_model_kv_override_free(struct llama_model_kv_override * kv_overrides) {
+    for (const struct llama_model_kv_override *p = kv_overrides; p->key[0] != 0; p++) {
+        if (p->tag == LLAMA_KV_OVERRIDE_TYPE_STR) {
+            delete p->str_value;
+        }
+    }
+}
+
 struct llama_context * llama_new_context_with_model(
                  struct llama_model * model,
         struct llama_context_params   params) {
diff --git a/llama.h b/llama.h
index 0cc0d3f29..0f215f336 100644
--- a/llama.h
+++ b/llama.h
@@ -209,6 +209,9 @@ extern "C" {
         };
     };
 
+    // Frees all allocated memory
+    LLAMA_API void llama_model_kv_override_free(struct llama_model_kv_override * ctx);
+
     struct llama_model_params {
         int32_t n_gpu_layers; // number of layers to store in VRAM
         enum llama_split_mode split_mode; // how to split the model across multiple GPUs