diff --git a/ggml/src/ggml-cuda.cu b/ggml/src/ggml-cuda.cu
index ed784ea1..39e345b6 100644
--- a/ggml/src/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda.cu
@@ -1876,7 +1876,8 @@ static void ggml_cuda_mul_mat(ggml_backend_cuda_context & ctx, const ggml_tensor
 
     bool use_dequantize_mul_mat_vec = (ggml_is_quantized(src0->type) || src0->type == GGML_TYPE_F16)
         && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32
-        && src0->ne[0] % GGML_CUDA_DMMV_X == 0 && src1->ne[1] == 1;
+        && src0->ne[0] % GGML_CUDA_DMMV_X == 0 && src0->ne[0] >= GGML_CUDA_DMMV_X*2
+        && src1->ne[1] == 1;
     bool          use_mul_mat_vec_q =  ggml_is_quantized(src0->type)
         && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32
         && src1->ne[1] <= MMVQ_MAX_BATCH_SIZE;
diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index f2c58c37..9e2b1d8c 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -19477,7 +19477,7 @@ void ggml_graph_dump_dot(const struct ggml_cgraph * gb, const struct ggml_cgraph
 
     fprintf(fp, "digraph G {\n");
     fprintf(fp, "  newrank = true;\n");
-    fprintf(fp, "  rankdir = LR;\n");
+    fprintf(fp, "  rankdir = TB;\n");
 
     for (int i = 0; i < gb->n_nodes; i++) {
         struct ggml_tensor * node = gb->nodes[i];
@@ -19539,7 +19539,7 @@ void ggml_graph_dump_dot(const struct ggml_cgraph * gb, const struct ggml_cgraph
         }
 
         fprintf(fp, "CONST %d [%" PRId64 ", %" PRId64 "]", i, node->ne[0], node->ne[1]);
-        if (ggml_nelements(node) < 5) {
+        if (ggml_nelements(node) < 5 && node->data != NULL) {
             fprintf(fp, " | (");
             for (int j = 0; j < ggml_nelements(node); j++) {
                 if (node->type == GGML_TYPE_I8 || node->type == GGML_TYPE_I16 || node->type == GGML_TYPE_I32) {