From: Johannes Gäßler <redacted>
Date: Fri, 16 Jun 2023 18:25:51 +0000 (+0200)
Subject: llama : fix embd when offloading non-repeating layers (#1891)
X-Git-Tag: gguf-v0.4.0~620
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=ac3b8869538c7fbdb48ff141d78c4dea091789f0;p=pkg%2Fggml%2Fsources%2Fllama.cpp

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

diff --git a/llama.cpp b/llama.cpp
index a9043884..81f047ed 100644
--- 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;