diff --git a/ggml.c b/ggml.c
index 7ea1abfcd..336c3c0ff 100644
--- a/ggml.c
+++ b/ggml.c
@@ -12272,7 +12272,7 @@ static void ggml_compute_forward_soft_max_f32(
 
     GGML_TENSOR_UNARY_OP_LOCALS
 
-    const int64_t ne11 = src1 ? src1->ne[1] : 1;
+    //const int64_t ne11 = src1 ? src1->ne[1] : 1;
 
     // TODO: is this supposed to be ceil instead of floor?
     //       https://huggingface.co/mosaicml/mpt-7b/blob/main/attention.py#L370
diff --git a/tests/test-backend-ops.cpp b/tests/test-backend-ops.cpp
index b8994cecf..50c4b2729 100644
--- a/tests/test-backend-ops.cpp
+++ b/tests/test-backend-ops.cpp
@@ -2261,7 +2261,7 @@ static bool test_backend(ggml_backend_t backend, test_mode mode, const char * op
     test_cases.emplace_back(new test_leaky_relu());
 
 #if 1
-    for (int hs : { 128, 64, 80, }) {
+    for (int hs : { 128, 256, 64, 80, }) {
         for (int nh : { 32, }) {
             for (int kv : { 512, 1024, 2048, 4096, }) {
                 for (int nb : { 1, 2, 4, 8, 512, 1024, 2048, }) {