llama : fix --mtest option (close #1414)

1 files changed, 1 insertions, 1 deletions

diff --git a/examples/main/main.cpp b/examples/main/main.cpp
index bd1c4ab..8543414 100644
--- a/examples/main/main.cpp
+++ b/examples/main/main.cpp
@@ -121,7 +121,7 @@ int main(int argc, char ** argv) {
     // uncomment the "used_mem" line in llama.cpp to see the results
     if (params.mem_test) {
         {
-            const std::vector<llama_token> tmp(params.n_batch, 0);
+            const std::vector<llama_token> tmp(params.n_batch, llama_token_bos());
             llama_eval(ctx, tmp.data(), tmp.size(), 0, params.n_threads);
         }