diff --git a/dsv4/kernels/router/dense_router_decode_kernel.py b/dsv4/kernels/router/dense_router_decode_kernel.py index 4941ff3b..31a5b326 100644 --- a/dsv4/kernels/router/dense_router_decode_kernel.py +++ b/dsv4/kernels/router/dense_router_decode_kernel.py @@ -101,8 +101,8 @@ class DenseRouterDecodeKernel: self.num_tmem_alloc_cols = utils.get_num_tmem_alloc_cols(tCtAcc_fake) def run(self, X, W_gate, e_bias, out_w, out_ids, M, E, K, scaling, top_k, stream=None): - self.a_major_mode = tcgen05.OperandMajorMode.MAJOR_K - self.b_major_mode = tcgen05.OperandMajorMode.MAJOR_K + self.a_major_mode = tcgen05.OperandMajorMode.K + self.b_major_mode = tcgen05.OperandMajorMode.K self._setup_attributes() X_cu = cutlass_torch.to_cuTe_tensor(X, major_mode=self.a_major_mode)