diff --git a/paddle/fluid/operators/bernoulli_op.cu b/paddle/fluid/operators/bernoulli_op.cu
index d0837071d456068f64ebc74b115f1a7904eba41c..f665d2dd0e991847de2ad35bf6b18741fb3a6e26 100644
--- a/paddle/fluid/operators/bernoulli_op.cu
+++ b/paddle/fluid/operators/bernoulli_op.cu
@@ -31,6 +31,10 @@ struct BernoulliCudaFunctor {
   __host__ __device__ BernoulliCudaFunctor(int seed) : seed_(seed) {}
 
   __host__ __device__ T operator()(const unsigned int n, const T p) const {
+    // NOTE(zhiqiu): currently, PADDLE_ENFORCE in cuda kernel may print several
+    // lines of error messages if, and it should be refined.
+    PADDLE_ENFORCE(p >= 0.0 && p <= 1.0,
+                   "The probability should be >=0 and <= 1, but got %f", p);
     thrust::minstd_rand rng;
     rng.seed(seed_);
     thrust::uniform_real_distribution<T> dist(0.0, 1.0);
diff --git a/paddle/fluid/operators/bernoulli_op.h b/paddle/fluid/operators/bernoulli_op.h
index 06a83ada17bb926d6f7d4eef10750986d00f048c..40f285d11f194057d950f45798bea07439398ab0 100644
--- a/paddle/fluid/operators/bernoulli_op.h
+++ b/paddle/fluid/operators/bernoulli_op.h
@@ -25,10 +25,12 @@ namespace operators {
 
 template <typename T>
 inline HOSTDEVICE T BernoulliFunctor(T p, T rand) {
-  PADDLE_ENFORCE_LE(p, 1, platform::errors::OutOfRange(
-                              "The probability should be <= 1, but got %f", p));
-  PADDLE_ENFORCE_GE(p, 0, platform::errors::OutOfRange(
-                              "The probability should be >= 1, but got %f", p));
+  PADDLE_ENFORCE_LE(p, 1.0,
+                    platform::errors::OutOfRange(
+                        "The probability should be <= 1, but got %f", p));
+  PADDLE_ENFORCE_GE(p, 0.0,
+                    platform::errors::OutOfRange(
+                        "The probability should be >= 0, but got %f", p));
   return static_cast<T>(rand < p);
 }
 
diff --git a/paddle/fluid/operators/unsqueeze_op.cc b/paddle/fluid/operators/unsqueeze_op.cc
index c33e7c6068648d019a38450a92fec79032411598..ee1361e3618302816200efc759ebd18ee05c9274 100644
--- a/paddle/fluid/operators/unsqueeze_op.cc
+++ b/paddle/fluid/operators/unsqueeze_op.cc
@@ -13,9 +13,11 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/fluid/operators/unsqueeze_op.h"
+
 #include <memory>
 #include <string>
 #include <vector>
+
 #include "paddle/fluid/framework/op_registry.h"
 
 namespace paddle {
@@ -327,6 +329,7 @@ REGISTER_OPERATOR(unsqueeze2_grad, ops::Unsqueeze2GradOp,
 REGISTER_OP_CPU_KERNEL(
     unsqueeze, ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, float>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, bool>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, int>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, int8_t>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, int64_t>);
@@ -334,12 +337,14 @@ REGISTER_OP_CPU_KERNEL(
     unsqueeze_grad,
     ops::UnsqueezeGradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::UnsqueezeGradKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::UnsqueezeGradKernel<paddle::platform::CPUDeviceContext, bool>,
     ops::UnsqueezeGradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::UnsqueezeGradKernel<paddle::platform::CPUDeviceContext, int8_t>,
     ops::UnsqueezeGradKernel<paddle::platform::CPUDeviceContext, int64_t>);
 REGISTER_OP_CPU_KERNEL(
     unsqueeze2, ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, float>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, bool>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, int>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, int8_t>,
     ops::UnsqueezeKernel<paddle::platform::CPUDeviceContext, int64_t>);
@@ -347,6 +352,7 @@ REGISTER_OP_CPU_KERNEL(
     unsqueeze2_grad,
     ops::Unsqueeze2GradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::Unsqueeze2GradKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::Unsqueeze2GradKernel<paddle::platform::CPUDeviceContext, bool>,
     ops::Unsqueeze2GradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::Unsqueeze2GradKernel<paddle::platform::CPUDeviceContext, int8_t>,
     ops::Unsqueeze2GradKernel<paddle::platform::CPUDeviceContext, int64_t>);
diff --git a/paddle/fluid/operators/unsqueeze_op.cu.cc b/paddle/fluid/operators/unsqueeze_op.cu.cc
index 3258de53b8b7cda994c9555bf6a62502f3c04c23..0e8f47a692380cc96a371bb7a5319af89a3d28c4 100644
--- a/paddle/fluid/operators/unsqueeze_op.cu.cc
+++ b/paddle/fluid/operators/unsqueeze_op.cu.cc
@@ -21,6 +21,7 @@ REGISTER_OP_CUDA_KERNEL(
     unsqueeze, ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, float>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, double>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, plat::float16>,
+    ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, bool>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, int>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, int8_t>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, int64_t>);
@@ -30,6 +31,7 @@ REGISTER_OP_CUDA_KERNEL(
     ops::UnsqueezeGradKernel<paddle::platform::CUDADeviceContext, double>,
     ops::UnsqueezeGradKernel<paddle::platform::CUDADeviceContext,
                              plat::float16>,
+    ops::UnsqueezeGradKernel<paddle::platform::CUDADeviceContext, bool>,
     ops::UnsqueezeGradKernel<paddle::platform::CUDADeviceContext, int>,
     ops::UnsqueezeGradKernel<paddle::platform::CUDADeviceContext, int8_t>,
     ops::UnsqueezeGradKernel<paddle::platform::CUDADeviceContext, int64_t>);
@@ -38,6 +40,7 @@ REGISTER_OP_CUDA_KERNEL(
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, float>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, double>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, plat::float16>,
+    ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, bool>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, int>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, int8_t>,
     ops::UnsqueezeKernel<paddle::platform::CUDADeviceContext, int64_t>);
@@ -47,6 +50,7 @@ REGISTER_OP_CUDA_KERNEL(
     ops::Unsqueeze2GradKernel<paddle::platform::CUDADeviceContext, double>,
     ops::Unsqueeze2GradKernel<paddle::platform::CUDADeviceContext,
                               plat::float16>,
+    ops::Unsqueeze2GradKernel<paddle::platform::CUDADeviceContext, bool>,
     ops::Unsqueeze2GradKernel<paddle::platform::CUDADeviceContext, int>,
     ops::Unsqueeze2GradKernel<paddle::platform::CUDADeviceContext, int8_t>,
     ops::Unsqueeze2GradKernel<paddle::platform::CUDADeviceContext, int64_t>);
diff --git a/python/paddle/tensor/manipulation.py b/python/paddle/tensor/manipulation.py
index 845d2cf4d199328bbf8d0e03cd3a7a24a61aafd2..5a01fff88c16bfa584479d71ea93d78999de40df 100644
--- a/python/paddle/tensor/manipulation.py
+++ b/python/paddle/tensor/manipulation.py
@@ -433,8 +433,8 @@ def stack(x, axis=0, name=None):
                           [5.0, 6.0] ] ]
 
     Args:
-        x (Tensor|list[Tensor]): Input ``x`` can be a single tensor, or a ``list`` of tensors.
-                                     If ``x`` is a ``list``, the Tensors in ``x``
+        x (Tensor|list[Tensor]|tuple[Tensor]): Input ``x`` can be a single tensor, or a ``list`` or ``tuple`` of tensors.
+                                     If ``x`` is a ``list`` or ``tuple`` , the Tensors in ``x``
                                      must be of the same shape and dtype. Supported data types: float32, float64, int32, int64.
         axis (int, optional): The axis along which all inputs are stacked. ``axis`` range is ``[-(R+1), R+1)``,
                               where ``R`` is the number of dimensions of the first input tensor ``x[0]``.