More correct load progress

3 years ago · 4a4718e8ab
parent 23035f9ba8
commit 4a4718e8ab
1 changed files with 1 additions and 1 deletions
--- a/llama.cpp
+++ b/llama.cpp
@ -592,7 +592,7 @@ static bool llama_model_load(
                //fprintf(stderr, "%42s - [%5d, %5d], type = %6s, %6.2f MB\n", name.data(), ne[0], ne[1], ftype == 0 ? "float" : "f16", ggml_nbytes(tensor)/1024.0/1024.0);
                if (++n_tensors % 8 == 0) {
                    if (progress_callback) {
-                        double current_progress = (double(i) + (double(fin.tellg()) / double(file_size))) / double(n_parts);
+                        double current_progress = (double(i) + (double(size_t(fin.tellg()) - file_offset) / double(file_size - file_offset))) / double(n_parts);
                        progress_callback(current_progress, progress_callback_user_data);
                    }
                    fprintf(stderr, ".");