llama : fix embd when offloading non-repeating layers (#1891)

2025-10-31 08:51:55 +00:00 · 2023-06-16 20:25:51 +02:00
parent 5b9ccaf104
commit ac3b886953
1 changed files with 1 additions and 1 deletions
--- a/llama.cpp
+++ b/llama.cpp
@@ -1658,7 +1658,7 @@ static bool llama_eval_internal(

        // cur = cur*norm(broadcasted)
        cur = ggml_mul(ctx0, cur, model.norm);
-        offload_func_nr(cur);
+        // offload_func_nr(cur); // TODO CPU + GPU mirrored backend
        ggml_set_name(cur, "result_norm");

        embeddings = cur;