flashinfer.comm.mnnvl_cutedsl_moe_finalize.supports_moe_finalize_rmsnorm_allreduce

flashinfer.comm.mnnvl_cutedsl_moe_finalize.supports_moe_finalize_rmsnorm_allreduce(workspace: MNNVLCuteDSLMoEFinalizeAllReduceWorkspace, routed_rows: Tensor, route_weights: Tensor, permuted_idx: Tensor, norm_weight: Tensor, output: Tensor) → bool

Check only this rank. Use workspace.check_support to agree across ranks.