File tree Expand file tree Collapse file tree 3 files changed +7
-7
lines changed Expand file tree Collapse file tree 3 files changed +7
-7
lines changed Original file line number Diff line number Diff line change @@ -1558,7 +1558,8 @@ int main(int argc, char ** argv) {
15581558 }
15591559 prev_inst = &inst;
15601560 } else {
1561- llama_model_reset_time (lmodel);
1561+ // ensure load_time dost not accumulate in llama_bench when not loading the same model
1562+ llama_reset_model_time (lmodel);
15621563 }
15631564
15641565 llama_context * ctx = llama_new_context_with_model (lmodel, inst.to_llama_cparams ());
Original file line number Diff line number Diff line change @@ -414,7 +414,7 @@ extern "C" {
414414 const char * path_model,
415415 struct llama_model_params params);
416416
417- LLAMA_API void llama_model_reset_time (struct llama_model * model);
417+ LLAMA_API void llama_reset_model_time (struct llama_model * model);
418418
419419 LLAMA_API void llama_free_model (struct llama_model * model);
420420
Original file line number Diff line number Diff line change @@ -8809,11 +8809,6 @@ static bool llm_load_tensors(
88098809 return true;
88108810}
88118811
8812- void llama_model_reset_time(llama_model * model) {
8813- model->t_start_us = ggml_time_us();
8814- model->t_load_us = ggml_time_us() - model->t_start_us;
8815- }
8816-
88178812// Returns 0 on success, -1 on error, and -2 on cancellation via llama_progress_callback
88188813static int llama_model_load(const std::string & fname, llama_model & model, llama_model_params & params) {
88198814 model.t_start_us = ggml_time_us();
@@ -18695,6 +18690,10 @@ struct llama_model * llama_load_model_from_file(
1869518690 return model;
1869618691}
1869718692
18693+ void llama_reset_model_time(llama_model * model) {
18694+ model->t_start_us = ggml_time_us() - model->t_load_us;
18695+ }
18696+
1869818697void llama_free_model(struct llama_model * model) {
1869918698 delete model;
1870018699}
You can’t perform that action at this time.
0 commit comments