Home Explore Help
Sign In
cturan
/
llama.cpp
mirror of https://github.com/cturan/llama.cpp
1
0
Fork 0
Files Issues 0 Wiki
Browse Source

metal : enable mat-vec kernels for bs <= 4 (#10491)

Georgi Gerganov 1 year ago
parent
80acb7b430
commit
106964e3d2
1 changed files with 1 additions and 1 deletions
Unified View Show Diff Stats
  1. 1 1
      ggml/src/ggml-metal/ggml-metal.m

+ 1 - 1
ggml/src/ggml-metal/ggml-metal.m
View File 

@@ -1927,7 +1927,7 @@ static void ggml_metal_encode_node(
 
 
 
                 // find the break-even point where the matrix-matrix kernel becomes more efficient compared
 
                 // find the break-even point where the matrix-matrix kernel becomes more efficient compared
 
                 // to the matrix-vector kernel
 
                 // to the matrix-vector kernel
 
-                int ne11_mm_min = 1;
 
+                int ne11_mm_min = 4;
 
 
 
 #if 0
 
 #if 0
 
                 // the numbers below are measured on M2 Ultra for 7B and 13B models
 
                 // the numbers below are measured on M2 Ultra for 7B and 13B models