fix: inccorect bindings for kv override. Based on #1011

Author: abetlen

llama_cpp/llama_cpp.py

@@ -9,13 +9,15 @@
     c_int32,
     c_uint8,
     c_uint32,
+    c_int64,
     c_size_t,
     c_float,
     c_double,
     c_void_p,
     POINTER,
     _Pointer,  # type: ignore
     Structure,
+    Union as CtypesUnion,
     Array,
 )
 import pathlib
@@ -317,12 +319,9 @@ class llama_batch(Structure):
 #     LLAMA_KV_OVERRIDE_FLOAT,
 #     LLAMA_KV_OVERRIDE_BOOL,
 # };
-class llama_model_kv_override_type(Structure):
-    _fields_ = [
-        ("LLAMA_KV_OVERRIDE_INT", c_int),
-        ("LLAMA_KV_OVERRIDE_FLOAT", c_int),
-        ("LLAMA_KV_OVERRIDE_BOOL", c_int),
-    ]
+LLAMA_KV_OVERRIDE_INT = 0
+LLAMA_KV_OVERRIDE_FLOAT = 1
+LLAMA_KV_OVERRIDE_BOOL = 2
 
 # struct llama_model_kv_override {
 #     char key[128];
@@ -333,15 +332,20 @@ class llama_model_kv_override_type(Structure):
 #         bool bool_value;
 #     };
 # };
-class llama_model_kv_override(Structure):
+class llama_model_kv_override_value(CtypesUnion):
     _fields_ = [
-        ("key", ctypes.c_char * 128),
-        ("tag", llama_model_kv_override_type),
-        ("int_value", ctypes.c_int64),
+        ("int_value", c_int64),
         ("float_value", c_double),
         ("bool_value", c_bool),
     ]
 
+class llama_model_kv_override(Structure):
+    _fields_ = [
+        ("key", ctypes.c_char * 128),
+        ("tag", c_int),
+        ("value", llama_model_kv_override_value),
+    ]
+
 # struct llama_model_params {
 #     int32_t n_gpu_layers; // number of layers to store in VRAM
 #     int32_t main_gpu;     // the GPU that is used for scratch and small tensors