diff --git a/docs/sample_code/distributed_gradient_accumulation/train.py b/docs/sample_code/distributed_gradient_accumulation/train.py
index 19e26209d51517c0228f477653796fdf6a5fed30..2d4cf7c9f39e06f9246f0f29806dc6ce6c6b85f7 100644
--- a/docs/sample_code/distributed_gradient_accumulation/train.py
+++ b/docs/sample_code/distributed_gradient_accumulation/train.py
@@ -20,7 +20,7 @@ import mindspore as ms
 import mindspore.dataset as ds
 from mindspore import nn, train
 from mindspore.communication import init
-from mindspore.parallel import GradAccumulation
+from mindspore.parallel.nn import GradAccumulation
 from mindspore.parallel.auto_parallel import AutoParallel
 from mindspore.nn.utils import no_init_parameters
 
diff --git a/tutorials/source_en/parallel/distributed_gradient_accumulation.md b/tutorials/source_en/parallel/distributed_gradient_accumulation.md
index 01f13cc5914efbf734a2fc29b4dd5cd2e2123c7d..1fe0a0a728bd9dddf84a0d83b5670139eb2e3596 100644
--- a/tutorials/source_en/parallel/distributed_gradient_accumulation.md
+++ b/tutorials/source_en/parallel/distributed_gradient_accumulation.md
@@ -117,7 +117,7 @@ In this step, we need to define the loss function and the training process. Para
 ```python
 import mindspore as ms
 from mindspore import nn, train
-from mindspore.parallel import GradAccumulation
+from mindspore.parallel.nn import GradAccumulation
 
 loss_fn = nn.CrossEntropyLoss()
 loss_cb = train.LossMonitor(100)
diff --git a/tutorials/source_zh_cn/parallel/distributed_gradient_accumulation.md b/tutorials/source_zh_cn/parallel/distributed_gradient_accumulation.md
index 14965bd6a173b056676ec0882cf96cf62c7ce083..e9d3993bddf3e47983eb387d950017867e349969 100644
--- a/tutorials/source_zh_cn/parallel/distributed_gradient_accumulation.md
+++ b/tutorials/source_zh_cn/parallel/distributed_gradient_accumulation.md
@@ -117,7 +117,7 @@ with no_init_parameters():
 ```python
 import mindspore as ms
 from mindspore import nn, train
-from mindspore.parallel import GradAccumulation
+from mindspore.parallel.nn import GradAccumulation
 
 loss_fn = nn.CrossEntropyLoss()
 loss_cb = train.LossMonitor(100)