model : set res->t_embd in SmallThinker models (#16782)

diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 2a83d66279b7912d238eed34553c2c68c0f1a443..a6a6fa7d7198ad22334a80e7d89408bc3509a750 100644 (file)
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -19339,6 +19339,7 @@ struct llm_build_smallthinker : public llm_graph_context{
 
         cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
         cb(cur, "result_norm", -1);
+        res->t_embd = cur;
 
         // lm_head
         cur = build_lora_mm(model.output, cur);