fix double-close of fout

Author: cebtenzzre

llama.cpp

@@ -4963,7 +4963,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
         gguf_add_tensor(ctx_out, meta);
     }
 
-    auto fout = llama_file(fname_out.c_str(), "wb");
+    llama_file fout(fname_out.c_str(), "wb");
 
     const size_t meta_size = gguf_get_meta_size(ctx_out);
 
@@ -5020,7 +5020,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
 
     auto big_loop = [
         &read_data_pool, &f32_conv_buf_pool, &work_pool, &ml, &quant_tensor_types, &quant_tensor_sizes, params, &pool,
-        nthread, &log_mutex, &hist_all, ctx_out, fout, &quant_tensor_offsets
+        nthread, &log_mutex, &hist_all, ctx_out, &fout, &quant_tensor_offsets
     ](int i, int thread_idx) {
         std::array<int64_t, 1 << 4> hist_cur = {};
         auto & read_data = read_data_pool[thread_idx];