llama : use LLM_KV_GENERAL_FILE_TYPE instead of gguf_find_key (#12672)

diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp
index 954ae65a37c9361e2dfafebef7a81ed3f08fe5df..047782e7d0fc8a7a0b4c789c8b8dd0c4545e8314 100644 (file)
--- a/src/llama-arch.cpp
+++ b/src/llama-arch.cpp
@@ -75,6 +75,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
     { LLM_KV_GENERAL_ARCHITECTURE,         "general.architecture"                  },
     { LLM_KV_GENERAL_QUANTIZATION_VERSION, "general.quantization_version"          },
     { LLM_KV_GENERAL_ALIGNMENT,            "general.alignment"                     },
+    { LLM_KV_GENERAL_FILE_TYPE,            "general.file_type"                     },
     { LLM_KV_GENERAL_NAME,                 "general.name"                          },
     { LLM_KV_GENERAL_AUTHOR,               "general.author"                        },
     { LLM_KV_GENERAL_VERSION,              "general.version"                       },

diff --git a/src/llama-arch.h b/src/llama-arch.h
index 7e2a3dd29f7afb3f0ff7db6f73620e139932b44d..297cfa4dae57190f38dc71c5ce8d6017ddb404f9 100644 (file)
--- a/src/llama-arch.h
+++ b/src/llama-arch.h
@@ -79,6 +79,7 @@ enum llm_kv {
     LLM_KV_GENERAL_ARCHITECTURE,
     LLM_KV_GENERAL_QUANTIZATION_VERSION,
     LLM_KV_GENERAL_ALIGNMENT,
+    LLM_KV_GENERAL_FILE_TYPE,
     LLM_KV_GENERAL_NAME,
     LLM_KV_GENERAL_AUTHOR,
     LLM_KV_GENERAL_VERSION,

diff --git a/src/llama-model-loader.cpp b/src/llama-model-loader.cpp
index 05d58ad90eba91ac302e1fe2f97f2cb60922fb71..1be0f2d6d6c20a852369e9df5779dfd85d228f06 100644 (file)
--- a/src/llama-model-loader.cpp
+++ b/src/llama-model-loader.cpp
@@ -640,9 +640,9 @@ llama_model_loader::llama_model_loader(
         ftype = (llama_ftype) (ftype | LLAMA_FTYPE_GUESSED);
 
         {
-            const int kid = gguf_find_key(meta.get(), "general.file_type"); // TODO: use LLM_KV
-            if (kid >= 0) {
-                ftype = (llama_ftype) gguf_get_val_u32(meta.get(), kid);
+            uint32_t ftype_val = 0;
+            if (get_key(LLM_KV_GENERAL_FILE_TYPE, ftype_val, false)) {
+                ftype = (llama_ftype) ftype_val;
             }
         }