Apply suggestions from code review

Author: ngxson

examples/batched-bench/batched-bench.cpp

@@ -147,7 +147,7 @@ int main(int argc, char ** argv) {
                     llama_batch_ext_clear(batch);
 
                     for (int j = 0; j < pl; ++j) {
-                        llama_batch_ext_add_text(batch, 0, pp + i, &j, 1, false);
+                        llama_batch_ext_add_text(batch, 0, pp + i, &j, 1, true);
                     }
 
                     if (!decode_helper(ctx, batch, ctx_params.n_batch)) {

examples/batched/batched.cpp

@@ -196,7 +196,7 @@ int main(int argc, char ** argv) {
             i_batch[i] = llama_batch_ext_get_n_tokens(batch);
 
             // push this new token for next evaluation
-            llama_batch_ext_add_text(batch, new_token_id, n_cur, &i, 1, false);
+            llama_batch_ext_add_text(batch, new_token_id, n_cur, &i, 1, true);
 
             n_decode += 1;
         }

examples/main/main.cpp

@@ -670,7 +670,6 @@ int main(int argc, char ** argv) {
                 LOG_DBG("eval: %s\n", string_from(ctx, embd).c_str());
 
                 auto batch = llama_batch_ext_ptr::init_from_text(&embd[i], n_eval, n_past, 0, true);
-                llama_batch_ext_set_output_last(batch.get());
                 if (llama_decode_ext(ctx, batch.get())) {
                     LOG_ERR("%s : failed to eval\n", __func__);
                     return 1;