diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index e77fa06e7..e0ea890b1 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -5843,18 +5843,18 @@ static void dequantize_row_q6_K_cuda(const void * vx, dst_t * y, const int k, cu
 #endif
 }
 
-template <typename src_t, typename dst_t>
-static void convert_unary_cuda(const void * __restrict__ vx, dst_t * __restrict__ y, const int k, cudaStream_t stream) {
-    const int num_blocks = (k + CUDA_DEQUANTIZE_BLOCK_SIZE - 1) / CUDA_DEQUANTIZE_BLOCK_SIZE;
-    convert_unary<src_t><<<num_blocks, CUDA_DEQUANTIZE_BLOCK_SIZE, 0, stream>>>(vx, y, k);
-}
-
 template<typename dst_t>
 static void dequantize_row_iq2_xxs_cuda(const void * vx, dst_t * y, const int k, cudaStream_t stream) {
     const int nb = k / QK_K;
     dequantize_block_iq2_xxs<<<nb, 32, 0, stream>>>(vx, y);
 }
 
+template <typename src_t, typename dst_t>
+static void convert_unary_cuda(const void * __restrict__ vx, dst_t * __restrict__ y, const int k, cudaStream_t stream) {
+    const int num_blocks = (k + CUDA_DEQUANTIZE_BLOCK_SIZE - 1) / CUDA_DEQUANTIZE_BLOCK_SIZE;
+    convert_unary<src_t><<<num_blocks, CUDA_DEQUANTIZE_BLOCK_SIZE, 0, stream>>>(vx, y, k);
+}
+
 static to_fp16_cuda_t ggml_get_to_fp16_cuda(ggml_type type) {
     switch (type) {
         case GGML_TYPE_Q4_0: