diff --git a/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamw.h b/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamw.h
index 43f45d5575..b46bda7428 100644
--- a/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamw.h
+++ b/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamw.h
@@ -5,7 +5,7 @@
 
 #include "core/common/common.h"
 #include "core/framework/op_kernel.h"
-
+#include "core/providers/cpu/tensor/utils.h"
 #include "orttraining/training_ops/cpu/optimizer/adamw/adamwbase.h"
 
 namespace onnxruntime {
diff --git a/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamwbase.h b/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamwbase.h
index 6b10828cc0..7ee1ebe4a5 100644
--- a/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamwbase.h
+++ b/orttraining/orttraining/training_ops/cpu/optimizer/adamw/adamwbase.h
@@ -5,7 +5,6 @@
 
 #include "core/common/common.h"
 #include "core/framework/op_kernel.h"
-#include "core/providers/cpu/tensor/utils.h"
 
 namespace onnxruntime {
 namespace contrib {