From 9d38e741ff3b0555a651936aee5f0b0f11570e66 Mon Sep 17 00:00:00 2001
From: Your <wangye@microsoft.com>
Date: Sat, 23 Mar 2024 03:51:29 +0000
Subject: [PATCH] lintrunner

---
 .../cuda/moe/cutlass_extensions/gemm/kernel/fpA_intB_gemm.h | 6 +++---
 .../contrib_ops/cuda/moe/cutlass_extensions/gemm_configs.h  | 5 ++++-
 .../cuda/moe/cutlass_extensions/weight_only_quant_op.h      | 5 ++++-
 3 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm/kernel/fpA_intB_gemm.h b/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm/kernel/fpA_intB_gemm.h
index df5d1fc16ee0e..39663f6f1bc6c 100644
--- a/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm/kernel/fpA_intB_gemm.h
+++ b/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm/kernel/fpA_intB_gemm.h
@@ -61,9 +61,9 @@ template <typename Mma_,                 ///! Threadblock-scoped matrix multiply
           typename Epilogue_,            ///! Epilogue
           typename ThreadblockSwizzle_,  ///! Threadblock swizzling function
           typename KernelArch,           ///! The Architecture this kernel is compiled for. Used since SIMT kernels lose
-                                         ///top-level
-                                         /// arch.
-          bool SplitKSerial              ///! If true, code supporting split-K via serial reduction is enabled.
+                                         /// top-level
+                                /// arch.
+          bool SplitKSerial  ///! If true, code supporting split-K via serial reduction is enabled.
           >
 struct GemmFpAIntB {
   using Mma = Mma_;
diff --git a/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm_configs.h b/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm_configs.h
index 9c2b90e8c5d2f..25f02af458390 100644
--- a/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm_configs.h
+++ b/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/gemm_configs.h
@@ -93,7 +93,10 @@ enum class EpilogueScheduleType {
         // architectures older than hopper, the epilogue is always performed by the same thread block as the main loop.
 };
 
-enum class ClusterShape { ClusterShape_1x1x1, ClusterShape_2x1x1, ClusterShape_1x2x1, ClusterShape_2x2x1 };
+enum class ClusterShape { ClusterShape_1x1x1,
+                          ClusterShape_2x1x1,
+                          ClusterShape_1x2x1,
+                          ClusterShape_2x2x1 };
 
 struct CutlassGemmConfig {
   CutlassTileConfig tile_config = CutlassTileConfig::ChooseWithHeuristic;
diff --git a/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/weight_only_quant_op.h b/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/weight_only_quant_op.h
index 75fbd87d0d13c..403221a956017 100644
--- a/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/weight_only_quant_op.h
+++ b/onnxruntime/contrib_ops/cuda/moe/cutlass_extensions/weight_only_quant_op.h
@@ -36,7 +36,10 @@
 
 namespace cutlass {
 
-enum class WeightOnlyQuantOp { UNDEFINED, PER_COLUMN_SCALE_ONLY, FINEGRAINED_SCALE_ONLY, FINEGRAINED_SCALE_AND_ZEROS };
+enum class WeightOnlyQuantOp { UNDEFINED,
+                               PER_COLUMN_SCALE_ONLY,
+                               FINEGRAINED_SCALE_ONLY,
+                               FINEGRAINED_SCALE_AND_ZEROS };
 
 constexpr bool isFinegrained(WeightOnlyQuantOp op) {
   return op == WeightOnlyQuantOp::FINEGRAINED_SCALE_AND_ZEROS || op == WeightOnlyQuantOp::FINEGRAINED_SCALE_ONLY;