diff --git a/paddle/fluid/operators/kldiv_loss_op.cc b/paddle/fluid/operators/kldiv_loss_op.cc
index a3254c51c231d316ca504ccbdc3f175f2274e1ef..be84b57c6f46b1f891bd68e1c0dade731fc0503a 100644
--- a/paddle/fluid/operators/kldiv_loss_op.cc
+++ b/paddle/fluid/operators/kldiv_loss_op.cc
@@ -48,7 +48,7 @@ class KLDivLossOp : public framework::OperatorWithKernel {
     if ("none" == reduction) {
       ctx->SetOutputDim("Loss", dim_x);
     } else {
-      ctx->SetOutputDim("Loss", framework::make_ddim({1}));
+      ctx->SetOutputDim("Loss", {1});
     }
   }
 
@@ -81,7 +81,7 @@ class KLDivLossOpMaker : public framework::OpProtoAndCheckerMaker {
         "The reduction type to apply to the output, available types "
         "are 'none' | 'batchmean' | 'mean' | 'sum', 'none' for no "
         "reduction, 'batchmean' for the sum of output divided by "
-        "batchmean size, 'mean' for the average valud of all output, "
+        "batch size, 'mean' for the average valud of all output, "
         "'sum' for the sum of the output.")
         .SetDefault("mean");
 
diff --git a/paddle/fluid/operators/kldiv_loss_op.h b/paddle/fluid/operators/kldiv_loss_op.h
index f262cfbb5fb337058cfe6c901efc2452148520d2..625e16e298d9f842fa621aca727c6df2cb045301 100644
--- a/paddle/fluid/operators/kldiv_loss_op.h
+++ b/paddle/fluid/operators/kldiv_loss_op.h
@@ -104,7 +104,8 @@ class KLDivLossGradKernel : public framework::OpKernel<T> {
 
     auto loss_grad_expand = loss_grad_t.broadcast(Array1(expand));
     auto grad_t = target_t * loss_grad_expand;
-    input_grad_t.device(place) = target_t.binaryExpr(grad_t, KLDivLossBackward<T>());
+    input_grad_t.device(place) =
+        target_t.binaryExpr(grad_t, KLDivLossBackward<T>());
 
     if ("mean" == reduction) {
       input_grad_t.device(place) = input_grad_t / static_cast<T>(numel);