llama : deci : support ffn-free with attention (#13296)

diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 774e343fb1f179403f8b3d47b2695dad0dd464ca..7b66e2a03e38ec96415f2a40df0378cb5fc03ed0 100644 (file)
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -4792,7 +4792,7 @@ struct llm_build_deci : public llm_graph_context {
             }
 
             // FFN-free layer of Llama-3_1-Nemotron-Ultra-253B
-            if (n_head == 0 && n_ff == 0) {
+            if (n_ff == 0) {
                 continue;
             }