diff --git a/src/layer/arm/multiheadattention_arm.cpp b/src/layer/arm/multiheadattention_arm.cpp
index 7ea0f62272b..15eca715699 100644
--- a/src/layer/arm/multiheadattention_arm.cpp
+++ b/src/layer/arm/multiheadattention_arm.cpp
@@ -44,7 +44,8 @@ MultiHeadAttention_arm::MultiHeadAttention_arm()
 int MultiHeadAttention_arm::create_pipeline(const Option& _opt)
 {
     Option opt = _opt;
-    opt.use_bf16_storage = false;
+    opt.use_fp16_storage &= support_fp16_storage;
+    opt.use_bf16_storage &= support_bf16_storage;
 
     {
         qk_softmax = ncnn::create_layer(ncnn::LayerType::Softmax);
@@ -224,7 +225,8 @@ int MultiHeadAttention_arm::create_pipeline(const Option& _opt)
 int MultiHeadAttention_arm::destroy_pipeline(const Option& _opt)
 {
     Option opt = _opt;
-    opt.use_bf16_storage = false;
+    opt.use_fp16_storage &= support_fp16_storage;
+    opt.use_bf16_storage &= support_bf16_storage;
 
     if (qk_softmax)
     {
@@ -286,7 +288,8 @@ int MultiHeadAttention_arm::forward(const std::vector<Mat>& bottom_blobs, std::v
     const Mat& attn_mask_blob = attn_mask ? bottom_blobs[bottom_blobs.size() - 1] : Mat();
 
     Option opt = _opt;
-    opt.use_bf16_storage = false;
+    opt.use_fp16_storage &= support_fp16_storage;
+    opt.use_bf16_storage &= support_bf16_storage;
 
     Mat attn_mask_blob_unpacked;
     if (attn_mask_blob.elempack != 1)