diff --git a/paddle/fluid/eager/pylayer/py_layer_node.cc b/paddle/fluid/eager/pylayer/py_layer_node.cc
index 5ac0fc4640eb0..1aacd69ecc32d 100644
--- a/paddle/fluid/eager/pylayer/py_layer_node.cc
+++ b/paddle/fluid/eager/pylayer/py_layer_node.cc
@@ -50,7 +50,7 @@ GradNodePyLayer::operator()(
   PADDLE_ENFORCE_EQ(ctx->forward_output_tensor_is_duplicable.size(),
                     grads.size(),
                     paddle::platform::errors::InvalidArgument(
-                        "%s's grad input size(%s) mast be equal with it's "
+                        "%s's grad input size(%s) must be equal with it's "
                         "forward's output size(%s).",
                         name(),
                         grads.size(),
diff --git a/python/paddle/autograd/saved_tensors_hooks.py b/python/paddle/autograd/saved_tensors_hooks.py
index 5d68a90e78c89..2839f7acafbe6 100644
--- a/python/paddle/autograd/saved_tensors_hooks.py
+++ b/python/paddle/autograd/saved_tensors_hooks.py
@@ -35,7 +35,7 @@ class saved_tensors_hooks:
             backward need use the saved inputs/outputs tensors. Then you can reload
             the tensor and return it to paddle framework. The input of `unpack_hook`
             is the information returned by `pack_hook`. The output of `unpack_hook`
-            is a tensor reloaded by the information, and the tensor mast has the same
+            is a tensor reloaded by the information, and the tensor must has the same
             content as the original tensor passed as input to the corresponding
             `pack_hook`.
 
diff --git a/python/paddle/base/layers/layer_function_generator.py b/python/paddle/base/layers/layer_function_generator.py
index d71316ba90dc6..2404ef286b1f2 100644
--- a/python/paddle/base/layers/layer_function_generator.py
+++ b/python/paddle/base/layers/layer_function_generator.py
@@ -329,7 +329,7 @@ def func(x, name=None):
                 and x.is_view_var
             ):
                 raise ValueError(
-                    'Sorry about what\'s happend. In to_static mode, {}\'s output variable {} is a viewed Tensor in dygraph. This will result in inconsistent calculation behavior between dynamic and static graphs. You mast find the location of the strided API be called, and call {} = {}.assign().'.format(
+                    'Sorry about what\'s happend. In to_static mode, {}\'s output variable {} is a viewed Tensor in dygraph. This will result in inconsistent calculation behavior between dynamic and static graphs. You must find the location of the strided API be called, and call {} = {}.assign().'.format(
                         inplace_op_type, x.name, x.name, x.nameb
                     )
                 )
diff --git a/python/paddle/utils/inplace_utils.py b/python/paddle/utils/inplace_utils.py
index 306c3fba5bf64..a5f30ab91daaa 100644
--- a/python/paddle/utils/inplace_utils.py
+++ b/python/paddle/utils/inplace_utils.py
@@ -39,7 +39,7 @@ def __impl__(*args, **kwargs):
                 for arg in args:
                     if hasattr(arg, "is_view_var") and arg.is_view_var:
                         raise ValueError(
-                            f'Sorry about what\'s happend. In to_static mode, {func.__name__}\'s output variable {arg.name} is a viewed Tensor in dygraph. This will result in inconsistent calculation behavior between dynamic and static graphs. You mast find the location of the strided API be called, and call {arg.name} = {arg.name}.assign().'
+                            f'Sorry about what\'s happend. In to_static mode, {func.__name__}\'s output variable {arg.name} is a viewed Tensor in dygraph. This will result in inconsistent calculation behavior between dynamic and static graphs. You must find the location of the strided API be called, and call {arg.name} = {arg.name}.assign().'
                         )
 
             origin_func = f"{func.__module__}.{origin_api_name}"