llama-bench : use two tokens in the warmup run for prompt evals (#3059)

1 files changed, 6 insertions, 1 deletions

diff --git a/examples/llama-bench/llama-bench.cpp b/examples/llama-bench/llama-bench.cpp
index 72a02507..dedaa34f 100644
--- a/examples/llama-bench/llama-bench.cpp
+++ b/examples/llama-bench/llama-bench.cpp
@@ -986,7 +986,12 @@ int main(int argc, char ** argv) {
         test t(inst, lmodel, ctx);
 
         // warmup run
-        test_gen(ctx, 1, 0, t.n_threads);
+        if (t.n_prompt > 0) {
+            test_prompt(ctx, std::min(2, t.n_batch), 0, t.n_batch, t.n_threads);
+        }
+        if (t.n_gen > 0) {
+            test_gen(ctx, 1, 0, t.n_threads);
+        }
 
         for (int i = 0; i < params.reps; i++) {
             uint64_t t_start = get_time_ns();