llama: initialize f16 tables in quantize c api.

Author: AsakusaRinne

examples/quantize/quantize.cpp

@@ -52,13 +52,6 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
-    // needed to initialize f16 tables
-    {
-        struct ggml_init_params params = { 0, NULL, false };
-        struct ggml_context * ctx = ggml_init(params);
-        ggml_free(ctx);
-    }
-
     // parse command line arguments
     const std::string fname_inp = argv[1];
     std::string fname_out;

llama.cpp

@@ -2198,6 +2198,12 @@ int llama_model_quantize(
   enum llama_ftype   ftype,
         int          nthread) {
     try {
+        // needed to initialize f16 tables
+        {
+            struct ggml_init_params params = { 0, NULL, false };
+            struct ggml_context * ctx = ggml_init(params);
+            ggml_free(ctx);
+        }
         llama_model_quantize_internal(fname_inp, fname_out, ftype, nthread);
         return 0;
     } catch (const std::string & err) {