From 38328bb5998f555285fd32bace3de99937785e60 Mon Sep 17 00:00:00 2001
From: Michael Podvitskiy <podvitskiymichael@gmail.com>
Date: Wed, 13 Mar 2024 21:48:37 +0100
Subject: [PATCH] fragmentation calculation fix

---
 llama.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llama.cpp b/llama.cpp
index c431f769e..a7d53d890 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -9037,7 +9037,7 @@ static int llama_decode_internal(
 
     // decide if we need to defrag the kv cache
     if (cparams.defrag_thold >= 0.0f) {
-        const float fragmentation = kv_self.n >= 128 ? 1.0f - float(kv_self.used + n_tokens_all)/float(kv_self.n) : 0.0f;
+        const float fragmentation = kv_self.n >= 128 ? 1.0f - float(kv_self.used)/float(kv_self.n) : 0.0f;
 
         // queue defragmentation for next llama_kv_cache_update
         if (fragmentation > cparams.defrag_thold) {