diff --git a/torch/csrc/distributed/c10d/cuda/cutlass/gemm/kernel/persistent_async_input_scheduler.cuh b/torch/csrc/distributed/c10d/cuda/cutlass/gemm/kernel/persistent_async_input_scheduler.cuh
--- a/torch/csrc/distributed/c10d/cuda/cutlass/gemm/kernel/persistent_async_input_scheduler.cuh
+++ b/torch/csrc/distributed/c10d/cuda/cutlass/gemm/kernel/persistent_async_input_scheduler.cuh
@@ -220,6 +220,12 @@
   can_implement(Arguments const& args) {
     return args.max_swizzle_size >= 1;
   }
 
+  CUTLASS_HOST_DEVICE
+  static bool
+  can_implement(Arguments const& args, KernelHardwareInfo const&) {
+    return can_implement(args);
+  }
+
   CUTLASS_HOST_DEVICE
   PersistentTileSchedulerSm90AsyncInput() { }