From 0e41b94f40e1d10893d6ac29c727482573ef1652 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Mon, 7 Aug 2023 10:43:06 +0800
Subject: [PATCH] improve detection for 70B.

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

diff --git a/llama.cpp b/llama.cpp
index e797d132f..d46bd31a9 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -1076,7 +1076,7 @@ static void llama_model_load_internal(
         // LLaMAv2
         // TODO: temporary until GGUF
         //patch for llama2 gqa
-        if (model.type == e_model::MODEL_65B && hparams.n_mult >= 4096) {
+        if (model.type == e_model::MODEL_65B && hparams.n_mult == 4096) {
             fprintf(stderr, "%s: Applying KCPP Patch for 70B model, setting GQA to 8\n", __func__);
             n_gqa = 8;
         }