From 3cb1c348b3481f39d244b6b0cd5afb04aa6bd460 Mon Sep 17 00:00:00 2001
From: Georgi Gerganov <ggerganov@gmail.com>
Date: Fri, 1 Dec 2023 21:47:42 +0200
Subject: [PATCH] metal : try to improve batched decoding

---
 ggml-metal.m | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ggml-metal.m b/ggml-metal.m
index 6cfacf64fcd9f..3343bc8a3af37 100644
--- a/ggml-metal.m
+++ b/ggml-metal.m
@@ -1083,7 +1083,7 @@ void ggml_metal_graph_compute(
 
                             // find the break-even point where the matrix-matrix kernel becomes more efficient compared
                             // to the matrix-vector kernel
-                            int ne11_mm_min = 1;
+                            int ne11_mm_min = src0t == GGML_TYPE_F16 ? 1 : 16;
 
 #if 0
                             // the numbers below are measured on M2 Ultra for 7B and 13B models