diff --git a/csrc/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu b/csrc/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu
index 54d85db8..46e3f858 100644
--- a/csrc/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu
+++ b/csrc/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu
@@ -73,7 +73,7 @@ void FMHACutlassSM100BwdRun(at::Tensor workspace_buffer, at::Tensor d_o, at::Ten
                           dq, dk, dv,
                           softmax_scale, max_seqlen_q, max_seqlen_kv);      }
       else {
-        std::cout << "No kernel instantiated for head_dim_qk=" << head_dim_qk << " head_dim_vo=" << head_dim_vo << std::endl;
+        TORCH_CHECK(false, "No kernel instantiated for head_dim_qk=", head_dim_qk, " head_dim_vo=", head_dim_vo);
       }
     });
 
diff --git a/csrc/sm100/prefill/dense/fmha_cutlass_fwd_sm100.cu b/csrc/sm100/prefill/dense/fmha_cutlass_fwd_sm100.cu
index ab66f0fd..c08fb145 100644
--- a/csrc/sm100/prefill/dense/fmha_cutlass_fwd_sm100.cu
+++ b/csrc/sm100/prefill/dense/fmha_cutlass_fwd_sm100.cu
@@ -72,8 +72,8 @@ void FMHACutlassSM100FwdRun(at::Tensor workspace_buffer, at::Tensor q, at::Tenso
                           cumulative_seqlen_q, cumulative_seqlen_kv, o, lse, sm_scale,
                           max_seqlen_q, max_seqlen_kv);
       } else {
-        std::cout << "No kernel instantiated for head_dim_qk=" << head_dim_qk
-                  << " head_dim_vo=" << head_dim_vo << std::endl;
+        TORCH_CHECK(false, "No kernel instantiated for head_dim_qk=", head_dim_qk,
+                    " head_dim_vo=", head_dim_vo);
       }
     });