docs: clarify split loading usage

2025-06-19 19:55:56 +08:00
parent 1e7ae71ce5
commit 6a898ef9d8
6 changed files with 91 additions and 25 deletions
@@ -1510,6 +1510,13 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
        }
    ).set_env("LLAMA_ARG_RPC"));
 #endif
+    add_opt(llama_arg(
+        {"--splits"}, "LIST",
+        "comma separated list of GGUF split indexes to load (add 0 to load tensors from the first split)",
+        [](gpt_params & params, const std::string & value) {
+            params.gguf_splits = value;
+        }
+    ));
    add_opt(llama_arg(
        {"--mlock"},
        "force system to keep model in RAM rather than swapping or compressing",