diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index d00f50a3..a3b36108 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -15021,7 +15021,10 @@ static void ggml_compute_forward_mul_mat_id_up_gate(
     if (ids->ne[1] == 1 && dst->type == GGML_TYPE_F32) {
         int gcd = simple_gcd(n_ids, nth);
         if (gcd > 1) {
-            ggml_barrier(params->shared);
+            if (src1->type != vec_dot_type) {
+                // make sure quantization has finished
+                ggml_barrier(params->shared);
+            }
             const void * wdata    = (src1->type == vec_dot_type) ? src1->data : params->wdata;
             const size_t row_size = ggml_row_size(vec_dot_type, ne10);
             int counter = 0;
diff --git a/ggml/src/iqk/iqk_mul_mat.cpp b/ggml/src/iqk/iqk_mul_mat.cpp
index 4d29e2f0..1242c2fb 100644
--- a/ggml/src/iqk/iqk_mul_mat.cpp
+++ b/ggml/src/iqk/iqk_mul_mat.cpp
@@ -532,7 +532,9 @@ bool iqk_moe_fused_up_gate(long Nx, long Ny, long ne00, int ne11, int unary_op,
         float * C, long nb1, long nb2, const void * vrow_mapping, int ith, int nth) {
 
     const mmid_row_mapping * row_mapping = (const mmid_row_mapping *)vrow_mapping;
-    assert(row_mapping != nullptr);
+    // Removing this assert to accomodate usage without row id mapping (e.g., for Ny = 1,
+    // or if B has been prepared to be contiguous.
+    //assert(row_mapping != nullptr);
 
     MulMat mm;
     if (!MulMat::prepare(typeA, typeB, ne00, mm, Ny)) {