diff --git a/tests/unit/test_umma_qk.cu b/tests/unit/test_umma_qk.cu
index cc5dd634..b4827cdf 100644
--- a/tests/unit/test_umma_qk.cu
+++ b/tests/unit/test_umma_qk.cu
@@ -62,23 +62,23 @@ test_umma_qk_hd16(
     }
 
     // ================================================================
-    // TMEM allocation
+    // TMEM allocation — SKIP for now
     // ================================================================
-    if (wid == 0) {
-        uint32_t smem_ptr = __cvta_generic_to_shared(sTmemBase);
-        tmem_alloc(smem_ptr, 128);
-    }
-    __syncthreads();
-    uint32_t tmem_base = *sTmemBase;
-
-    // Zero TMEM
-    if (wid == 0) {
-        for (int col = 0; col < 128; col++) {
-            tmem_store(tmem_base + col, 0, 0, 0, 0);
-        }
-        tmem_fence_store();
-    }
-    __syncthreads();
+    // if (wid == 0) {
+    //     uint32_t smem_ptr = __cvta_generic_to_shared(sTmemBase);
+    //     tmem_alloc(smem_ptr, 128);
+    // }
+    // __syncthreads();
+    // uint32_t tmem_base = *sTmemBase;
+    //
+    // // Zero TMEM
+    // if (wid == 0) {
+    //     for (int col = 0; col < 128; col++) {
+    //         tmem_store(tmem_base + col, 0, 0, 0, 0);
+    //     }
+    //     tmem_fence_store();
+    // }
+    // __syncthreads();
 
     // ================================================================
     // Load Q and K into SMEM in canonical layout
@@ -151,10 +151,10 @@ test_umma_qk_hd16(
     }
     __syncthreads();
 
-    // TMEM dealloc
-    if (wid == 0) {
-        tmem_dealloc(tmem_base, 128);
-    }
+    // TMEM dealloc — skipped
+    // if (wid == 0) {
+    //     tmem_dealloc(tmem_base, 128);
+    // }
 }
 
 // ==================================================================