diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index b82069589..a18295ea7 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -2,7 +2,7 @@
 #include "ggml.h"
 #include "ggml-backend-impl.h"
 
-static bool g_mul_mat_q = false;
+bool g_mul_mat_q = false;
 
 #include "ggml-cuda/common.cuh"
 #include "ggml-cuda/acc.cuh"
diff --git a/ggml-cuda/mmq.cuh b/ggml-cuda/mmq.cuh
index 31fcbf139..d59c0532f 100644
--- a/ggml-cuda/mmq.cuh
+++ b/ggml-cuda/mmq.cuh
@@ -2608,3 +2608,5 @@ void ggml_cuda_op_mul_mat_q(
     const int64_t src1_padded_row_size, cudaStream_t stream);
 
 bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11);
+
+extern bool g_mul_mat_q;
\ No newline at end of file