[cherry-pick 2.5][Zero-Dim] update 0D tensor API en doc (#53855)

* [Zero-Dim] update 0d tensor api en doc, test=document_fix

* [BUG] fix windows kernel dispatch of _lzcnt bug (#53728)

paddle/phi/api/lib/kernel_dispatch.cc

@@ -78,7 +78,10 @@ std::size_t CountLeadingZeros(uint32_t val) {
 #if defined(__clang__) || defined(__GNUC__)
   return __builtin_clz(val);
 #elif defined(_MSC_VER)
-  return __lzcnt(val);
+  // windows don't have built-in clz/ctz function
+  DWORD Index;
+  _BitScanReverse(&Index, val);
+  return (uint32_t)Index ^ 31;
 #else
   if (val == 0) {
     return 32;

paddle/phi/kernels/gpu/randperm_kernel.cu

@@ -35,8 +35,6 @@ namespace cub = hipcub;
 #include "paddle/phi/kernels/funcs/for_range.h"
 #include "paddle/phi/kernels/randint_kernel.h"
 
-DECLARE_bool(use_curand);
-
 namespace phi {
 
 template <typename keyT, typename dataT>

paddle/scripts/paddle_build.bat

@@ -211,7 +211,7 @@ if "%WITH_SCCACHE%"=="ON" (
 :install_sccache
 echo There is not sccache in this PC, will install sccache.
 echo Download package from https://paddle-ci.gz.bcebos.com/window_requirement/sccache.exe
-%PYTHON_ROOT%\python.exe -c "import wget;wget.download('https://paddle-ci.gz.bcebos.com/window_requirement/sccache.exe')"
+python -c "import wget;wget.download('https://paddle-ci.gz.bcebos.com/window_requirement/sccache.exe')"
 xcopy sccache.exe %SCCACHE_ROOT%\ /Y
 del sccache.exe
 goto:eof
@@ -429,7 +429,7 @@ if not exist %THIRD_PARTY_PATH% (
     if not exist %THIRD_PARTY_HOME% mkdir "%THIRD_PARTY_HOME%"
     cd /d %THIRD_PARTY_HOME%
     echo Getting third party: downloading ...
-    %PYTHON_ROOT%\python.exe -c "import wget;wget.download('https://paddle-windows.bj.bcebos.com/third_party/%sub_dir%/%md5%.tar.gz')" 2>nul
+    python -c "import wget;wget.download('https://paddle-windows.bj.bcebos.com/third_party/%sub_dir%/%md5%.tar.gz')" 2>nul
     if !ERRORLEVEL! EQU 0 (
         echo Getting third party: extracting ...
         tar -xf %md5%.tar.gz
@@ -572,17 +572,15 @@ if %ERRORLEVEL% NEQ 0 (
 )
 
 if "%UPLOAD_TP_FILE%"=="ON" (
-    set BCE_FILE=%cache_dir%\bce-python-sdk-0.8.33\BosClient.py
+    set BCE_FILE=%cache_dir%\bce-python-sdk-new\BosClient.py
     echo Uploading third_party: checking bce ...
-    if not exist %cache_dir%\bce-python-sdk-0.8.33 (
+    if not exist %cache_dir%\bce-python-sdk-new (
         echo There is no bce in this PC, will install bce.
         cd /d %cache_dir%
-        echo Download package from https://paddle-windows.bj.bcebos.com/bce-python-sdk-0.8.33.tar.gz
-        %PYTHON_ROOT%\python.exe -c "import wget;wget.download('https://paddle-windows.bj.bcebos.com/bce-python-sdk-0.8.33.tar.gz')"
-        %PYTHON_ROOT%\python.exe -c "import shutil;shutil.unpack_archive('bce-python-sdk-0.8.33.tar.gz', extract_dir='./',format='gztar')"
-        cd /d %cache_dir%\bce-python-sdk-0.8.33
-        %PYTHON_ROOT%\python.exe setup.py install 1>nul
-        del %cache_dir%\bce-python-sdk-0.8.33.tar.gz
+        echo Download package from https://xly-devops.bj.bcebos.com/home/bos_new.tar.gz
+        python -c "import wget;wget.download('https://xly-devops.bj.bcebos.com/home/bos_new.tar.gz')"
+        python -c "import shutil;shutil.unpack_archive('bos_new.tar.gz', extract_dir='./bce-python-sdk-new',format='gztar')"
+        python -m pip install bce-python-sdk
     )
     if !errorlevel! EQU 0 (
         cd /d %THIRD_PARTY_HOME%
@@ -590,7 +588,7 @@ if "%UPLOAD_TP_FILE%"=="ON" (
         tar -zcf %md5%.tar.gz %md5%
         if !errorlevel! EQU 0 (
             echo Uploading third_party: uploading ...
-            %PYTHON_ROOT%\python.exe !BCE_FILE! %md5%.tar.gz paddle-windows/third_party/%sub_dir% 1>nul
+            python !BCE_FILE! %md5%.tar.gz paddle-windows/third_party/%sub_dir% 1>nul
             if !errorlevel! EQU 0 (
                 echo Upload third party %md5% to bos paddle-windows/third_party/%sub_dir% successfully.
             ) else (

python/paddle/distributed/auto_parallel/operators/common.py

@@ -393,7 +393,7 @@ def get_data_parallel_group(dist_ctx, op, act_grad_names, rank):
 
     for var_name in act_grad_names:
         var_dim_mapping = op_dist_attr.get_input_dims_mapping(var_name)
-        # consider that the variable's shape is [], which is 0D
+        # consider that the variable's shape is [], which is 0-D
         # TODO utilize the batch_dim attr instead of "0" in future
         batch_size_axis = var_dim_mapping[0] if len(var_dim_mapping) > 0 else -1
 

python/paddle/distributed/communication/stream/all_gather.py

@@ -108,7 +108,7 @@ def _all_gather_in_static_mode(tensor_list, tensor, group, sync_op):
         },
     )
     tensor_list.clear()
-    # 0D use stack/unstack while others use concat/split
+    # 0-D use stack/unstack while others use concat/split
     if len(tensor.shape) == 0:
         tensor_list.extend(paddle.unstack(out, 0))
     else:

python/paddle/distributed/communication/stream/all_to_all.py

@@ -78,7 +78,7 @@ def _all_to_all_in_static_mode(
     if isinstance(in_tensor_or_tensor_list, list):
         if len(in_tensor_or_tensor_list) == 0:
             raise RuntimeError("The input tensor_list should not be empty.")
-        # 0D use stack/unstack while others use concat/split
+        # 0-D use stack/unstack while others use concat/split
         if len(in_tensor_or_tensor_list[0].shape) == 0:
             in_tensor = paddle.stack(in_tensor_or_tensor_list, axis=0)
         else:
@@ -115,7 +115,7 @@ def _all_to_all_in_static_mode(
     if isinstance(out_tensor_or_tensor_list, list):
         if not sync_op:
             dist.wait(out_tensor, use_calc_stream=False)
-        # 0D use stack/unstack while others use concat/split
+        # 0-D use stack/unstack while others use concat/split
         if len(in_tensor_or_tensor_list[0].shape) == 0:
             out_tensor_or_tensor_list.extend(paddle.unstack(out_tensor, 0))
         else:

python/paddle/distributed/communication/stream/scatter.py

@@ -91,7 +91,7 @@ def _scatter_in_static_mode(
                 )
         else:
             tensor_list = [tensor for _ in range(nranks)]
-        # 0D use stack/unstack while others use concat/split
+        # 0-D use stack/unstack while others use concat/split
         if len(tensor_list[0].shape) == 0:
             input_tensor = paddle.stack(tensor_list, axis=0)
         else:

python/paddle/distribution/bernoulli.py

@@ -79,16 +79,16 @@ class Bernoulli(exponential_family.ExponentialFamily):
             rv = Bernoulli(probs=0.3)
 
             print(rv.mean)
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [0.30000001])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        0.30000001)
 
             print(rv.variance)
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [0.21000001])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        0.21000001)
 
             print(rv.entropy())
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [0.61086434])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        0.61086434)
     """
 
     def __init__(self, probs, name=None):
@@ -247,12 +247,12 @@ class Bernoulli(exponential_family.ExponentialFamily):
 
                 # The smaller the `temperature`, the distribution of `rsample` closer to `sample`, with `probs` of 0.3.
                 print(paddle.nn.functional.sigmoid(rv.rsample([1000, ], temperature=1.0)).sum())
-                # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                #        [361.06829834])
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        361.06829834)
 
                 print(paddle.nn.functional.sigmoid(rv.rsample([1000, ], temperature=0.1)).sum())
-                # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                #        [288.66418457])
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        288.66418457)
         """
         name = self.name + '_rsample'
         if not _non_static_mode():
@@ -420,8 +420,8 @@ class Bernoulli(exponential_family.ExponentialFamily):
 
                 rv = Bernoulli(0.3)
                 print(rv.entropy())
-                # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                #        [0.61086434])
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        0.61086434)
         """
         name = self.name + '_entropy'
 
@@ -455,8 +455,8 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 rv_other = Bernoulli(0.7)
 
                 print(rv.kl_divergence(rv_other))
-                # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                #        [0.33891910])
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        0.33891910)
         """
         name = self.name + '_kl_divergence'
         if not _non_static_mode():

python/paddle/distribution/beta.py

@@ -61,13 +61,13 @@ class Beta(exponential_family.ExponentialFamily):
             beta = paddle.distribution.Beta(alpha=0.5, beta=0.5)
             print(beta.mean)
             # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #        [0.50000000])
+            #        0.50000000)
             print(beta.variance)
             # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #        [0.12500000])
+            #        0.12500000)
             print(beta.entropy())
             # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #        [0.12500000])
+            #        0.12500000)
 
             # tensor input with broadcast
             beta = paddle.distribution.Beta(alpha=paddle.to_tensor([0.2, 0.4]), beta=0.6)

python/paddle/distribution/dirichlet.py

@@ -63,10 +63,10 @@ class Dirichlet(exponential_family.ExponentialFamily):
 
             print(dirichlet.entropy())
             # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #        [-1.24434423])
+            #        -1.24434423)
             print(dirichlet.prob(paddle.to_tensor([.3, .5, .6])))
             # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #        [10.80000114])
+            #        10.80000114)
 
     """
 

python/paddle/distribution/gumbel.py

@@ -61,7 +61,7 @@ class Gumbel(TransformedDistribution):
           dist.cdf(value)
           # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True, [0.54523915])
           dist.entropy()
-          # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True, [1.57721567])
+          # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True, [1.57721567])
           dist.rsample([2])
           # Tensor(shape=[2, 1], dtype=float32, place=Place(gpu:0), stop_gradient=True, [[0.80463481], [0.91893655]])
 

python/paddle/distribution/independent.py

@@ -44,8 +44,8 @@ class Independent(distribution.Distribution):
             print(reinterpreted_beta.batch_shape, reinterpreted_beta.event_shape)
             # () (2,)
             print(reinterpreted_beta.log_prob(paddle.to_tensor([0.2,  0.2])))
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [-0.45687842])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        -0.45687842)
     """
 
     def __init__(self, base, reinterpreted_batch_rank):

python/paddle/distribution/kl.py

@@ -58,7 +58,7 @@ def kl_divergence(p, q):
 
             print(paddle.distribution.kl_divergence(p, q))
             # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #        [0.21193528])
+            #        0.21193528)
 
     """
     return _dispatch(type(p), type(q))(p, q)

python/paddle/distribution/laplace.py

@@ -44,12 +44,12 @@ class Laplace(distribution.Distribution):
     Examples:
         .. code-block:: python
 
-                        import paddle
+            import paddle
 
-                        m = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
-                        m.sample()  # Laplace distributed with loc=0, scale=1
-                        # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
-                        # [3.68546247])
+            m = paddle.distribution.Laplace(paddle.to_tensor(0.0), paddle.to_tensor(1.0))
+            m.sample()  # Laplace distributed with loc=0, scale=1
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        3.68546247)
 
     """
 
@@ -173,13 +173,13 @@ class Laplace(distribution.Distribution):
         Examples:
             .. code-block:: python
 
-                            import paddle
+                import paddle
 
-                            m = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
-                            value = paddle.to_tensor([0.1])
-                            m.log_prob(value)
-                            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                            # [-0.79314721])
+                m = paddle.distribution.Laplace(paddle.to_tensor(0.0), paddle.to_tensor(1.0))
+                value = paddle.to_tensor(0.1)
+                m.log_prob(value)
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        -0.79314721)
 
         """
         loc, scale, value = self._validate_value(value)
@@ -205,12 +205,12 @@ class Laplace(distribution.Distribution):
         Examples:
             .. code-block:: python
 
-                            import paddle
+                import paddle
 
-                            m = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
-                            m.entropy()
-                            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
-                            # [1.69314718])
+                m = paddle.distribution.Laplace(paddle.to_tensor(0.0), paddle.to_tensor(1.0))
+                m.entropy()
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        1.69314718)
         """
         return 1 + paddle.log(2 * self.scale)
 
@@ -236,13 +236,13 @@ class Laplace(distribution.Distribution):
         Examples:
             .. code-block:: python
 
-                            import paddle
+                import paddle
 
-                            m = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
-                            value = paddle.to_tensor([0.1])
-                            m.cdf(value)
-                            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                            # [0.54758132])
+                m = paddle.distribution.Laplace(paddle.to_tensor(0.0), paddle.to_tensor(1.0))
+                value = paddle.to_tensor(0.1)
+                m.cdf(value)
+                # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                #        0.54758132)
         """
         loc, scale, value = self._validate_value(value)
         iterm = (
@@ -277,11 +277,11 @@ class Laplace(distribution.Distribution):
 
                             import paddle
 
-                            m = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
-                            value = paddle.to_tensor([0.1])
+                            m = paddle.distribution.Laplace(paddle.to_tensor(0.0), paddle.to_tensor(1.0))
+                            value = paddle.to_tensor(0.1)
                             m.icdf(value)
-                            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-                            # [-1.60943794])
+                            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+                            #        -1.60943794)
         """
         loc, scale, value = self._validate_value(value)
         term = value - 0.5
@@ -302,10 +302,10 @@ class Laplace(distribution.Distribution):
 
                             import paddle
 
-                            m = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
+                            m = paddle.distribution.Laplace(paddle.to_tensor(0.0), paddle.to_tensor(1.0))
                             m.sample()  # Laplace distributed with loc=0, scale=1
                             # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
-                            # [3.68546247])
+                            #        3.68546247)
         """
         shape = shape if isinstance(shape, tuple) else tuple(shape)
         with paddle.no_grad():
@@ -395,13 +395,13 @@ class Laplace(distribution.Distribution):
         Examples:
             .. code-block:: python
 
-                            import paddle
+                import paddle
 
-                            m1 = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
-                            m2 = paddle.distribution.Laplace(paddle.to_tensor([1.0]), paddle.to_tensor([0.5]))
-                            m1.kl_divergence(m2)
-                            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
-                            # [1.04261160])
+                m1 = paddle.distribution.Laplace(paddle.to_tensor([0.0]), paddle.to_tensor([1.0]))
+                m2 = paddle.distribution.Laplace(paddle.to_tensor([1.0]), paddle.to_tensor([0.5]))
+                m1.kl_divergence(m2)
+                # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
+                # [1.04261160])
         """
 
         var_ratio = other.scale / self.scale

python/paddle/distribution/lognormal.py

@@ -72,13 +72,13 @@ class LogNormal(TransformedDistribution):
           sample = lognormal_a.sample((2, ))
           # a random tensor created by lognormal distribution with shape: [2, 1]
           entropy = lognormal_a.entropy()
-          # [1.4189385] with shape: []
+          # [1.4189385] with shape: [1]
           lp = lognormal_a.log_prob(value_tensor)
           # [-0.72069150] with shape: [1]
           p = lognormal_a.probs(value_tensor)
           # [0.48641577] with shape: [1]
           kl = lognormal_a.kl_divergence(lognormal_b)
-          # [0.34939718] with shape: []
+          # [0.34939718] with shape: [1]
     """
 
     def __init__(self, loc, scale):

python/paddle/distribution/normal.py

@@ -77,13 +77,13 @@ class Normal(distribution.Distribution):
             sample = normal_a.sample([2])
             # a random tensor created by normal distribution with shape: [2, 1]
             entropy = normal_a.entropy()
-            # [1.4189385] with shape: []
+            # [1.4189385] with shape: [1]
             lp = normal_a.log_prob(value_tensor)
             # [-1.2389386] with shape: [1]
             p = normal_a.probs(value_tensor)
             # [0.28969154] with shape: [1]
             kl = normal_a.kl_divergence(normal_b)
-            # [0.34939718] with shape: []
+            # [0.34939718] with shape: [1]
     """
 
     def __init__(self, loc, scale, name=None):

python/paddle/distribution/transform.py

@@ -435,8 +435,8 @@ class AffineTransform(Transform):
             # Tensor(shape=[2], dtype=float32, place=Place(gpu:0), stop_gradient=True,
             #        [1., 2.])
             print(affine.forward_log_det_jacobian(x))
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.)
     """
     _type = Type.BIJECTION
 
@@ -1189,8 +1189,8 @@ class StickBreakingTransform(Transform):
             # Tensor(shape=[3], dtype=float32, place=Place(gpu:0), stop_gradient=True,
             #        [0.99999988, 2.        , 2.99999881])
             print(t.forward_log_det_jacobian(x))
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [-9.10835075])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        -9.10835075)
     """
 
     _type = Type.BIJECTION

python/paddle/distribution/transformed_distribution.py

@@ -42,8 +42,8 @@ class TransformedDistribution(distribution.Distribution):
             #        [-0.10697651,  3.33609009, -0.86234951,  5.07457638,  0.75925219,
             #         -4.17087793,  2.22579336, -0.93845034,  0.66054249,  1.50957513])
             print(d.log_prob(paddle.to_tensor(0.5)))
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [-1.64333570])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        -1.64333570)
     """
 
     def __init__(self, base, transforms):

python/paddle/distribution/uniform.py

@@ -84,7 +84,7 @@ class Uniform(distribution.Distribution):
             sample = uniform.sample([2])
             # a random tensor created by uniform distribution with shape: [2, 1]
             entropy = uniform.entropy()
-            # [0.6931472] with shape: []
+            # [0.6931472] with shape: [1]
             lp = uniform.log_prob(value_tensor)
             # [-0.6931472] with shape: [1]
             p = uniform.probs(value_tensor)

python/paddle/fluid/backward.py

@@ -385,7 +385,7 @@ def _create_op_desc_(op_type, inputs, outputs, attrs):
 
 
 def _create_loss_op_desc_(loss):
-    # 0D Tensor or 0-Size Tensor
+    # 0-D Tensor or 0-Size Tensor
     if len(loss.shape) == 0 or 0 in loss.shape:
         create_shape = loss.shape
     else:

python/paddle/fluid/dygraph/learning_rate_scheduler.py

@@ -789,7 +789,7 @@ class ReduceLROnPlateau(LearningRateDecay):
     Reduce learning rate when ``loss`` has stopped descending. Models often benefit from reducing the learning rate
     by 2 to 10 times once model performance has no longer improvement.
 
-    The ``loss`` is the one which has been pass into ``step`` , it must be 1-D Tensor with shape [1]. When ``loss``
+    The ``loss`` is the one which has been pass into ``step`` , it must be 0-D Tensor with shape []. When ``loss``
     stop descending for a ``patience`` number of epochs, the learning rate will be reduced to ``learning_rate * decay_rate`` .
     (Specially, ``mode`` can also be set to ``'max`` , in this case, when ``loss`` stop ascending for a ``patience`` number
     of epochs, the learning rate will be reduced.)
@@ -943,7 +943,7 @@ class ReduceLROnPlateau(LearningRateDecay):
         Args:
             loss (Variable): A ``Variable`` that will be monitored to determine whether the learning rate will reduce.
                 If it stop descending for a ``patience`` number of epochs, the learning rate will reduce. It should
-                be 1-D Tensor with shape [1].
+                be 0-D Tensor with shape [].
                 Specially, if ``mode`` has been set to ``'max'`` ,  the learning rate will reduce when it stops ascending.
         Returns:
             None
@@ -952,7 +952,7 @@ class ReduceLROnPlateau(LearningRateDecay):
             Please refer to the example of current LearningRateDecay.
         """
 
-        # loss must be 1-D Tensor with shape [1]
+        # loss.size must be 1
         check_type(loss, 'loss', Variable, 'ReduceLROnPlateau.step')
         assert np.prod(loss.shape) == 1, (
             "The number of elements of loss should be 1, but the current loss.shape is {}, whose number of elements is not 1. "

python/paddle/fluid/dygraph/math_op_patch.py

@@ -156,7 +156,7 @@ def monkey_patch_math_varbase():
         return int(np.array(var).flatten()[0])
 
     def _len_(var):
-        assert var.ndim > 0, "len() of a 0D tensor is wrong"
+        assert var.ndim > 0, "len() of a 0-D tensor is wrong"
         if var.type == core.VarDesc.VarType.VOCAB:
             return len(var.value().get_map_tensor())
         elif var.type == core.VarDesc.VarType.STRINGS:

python/paddle/fluid/dygraph/varbase_patch_methods.py

@@ -569,7 +569,7 @@ def monkey_patch_varbase():
                 y = paddle.pow(x, 4.0)
                 y.backward()
                 print("grad of x: {}".format(x.grad))
-                # Tensor(shape=[1], dtype=float32, place=CUDAPlace(0), stop_gradient=False, [500.])
+                # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=False, 500.)
 
         """
         msg = (
@@ -688,12 +688,12 @@ def monkey_patch_varbase():
                 y = copy.deepcopy(x)
 
                 print(x)
-                # Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True,
-                #        [2.])
+                # Tensor(shape=[], dtype=float32, place=CPUPlace, stop_gradient=True,
+                #        2.)
 
                 print(y)
-                # Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True,
-                #        [2.])
+                # Tensor(shape=[], dtype=float32, place=CPUPlace, stop_gradient=True,
+                #        2.)
 
         """
         if not self.is_leaf:

python/paddle/fluid/framework.py

@@ -2512,7 +2512,7 @@ class Variable(metaclass=VariableMetaClass):
     def size(self):
         """
 
-        Returns the number of elements for current Variable, which is a int64 Variable with shape [1]
+        Returns the number of elements for current Variable, which is a int64 Variable with shape [] .
 
         Returns:
             Variable, the number of elements for current Variable

python/paddle/fluid/tests/unittests/test_numel_op.py

@@ -120,7 +120,6 @@ class TestNumelAPI(unittest.TestCase):
                 },
                 fetch_list=[out_1, out_2],
             )
-            # TODO(zhouwei): will change shape [1] to [] to support zero-dim
             assert np.array_equal(
                 res_1, np.array(np.size(input_1)).astype("int64")
             )

python/paddle/fluid/tests/unittests/test_size_op.py

@@ -83,7 +83,6 @@ class TestSizeAPI(unittest.TestCase):
                 },
                 fetch_list=[out_1, out_2],
             )
-            # TODO(zhouwei): will change shape [1] to [] to support zero-dim
             assert np.array_equal(
                 res_1, np.array(np.size(input_1)).astype("int64")
             )

python/paddle/jit/dy2static/convert_operators.py

@@ -517,7 +517,7 @@ def convert_len(var):
           `shape_op` in var.block.
     """
     if isinstance(var, Variable):
-        assert var.ndim > 0, "len() of a 0D tensor is wrong"
+        assert var.ndim > 0, "len() of a 0-D tensor is wrong"
         if var.type in [
             core.VarDesc.VarType.LOD_TENSOR,
             core.VarDesc.VarType.SELECTED_ROWS,

python/paddle/metric/metrics.py

@@ -798,7 +798,7 @@ def accuracy(input, label, k=1, correct=None, total=None, name=None):
             predictions = paddle.to_tensor([[0.2, 0.1, 0.4, 0.1, 0.1], [0.2, 0.3, 0.1, 0.15, 0.25]], dtype='float32')
             label = paddle.to_tensor([[2], [0]], dtype="int64")
             result = paddle.metric.accuracy(input=predictions, label=label, k=1)
-            # [0.5]
+            # 0.5
     """
     if label.dtype == paddle.int32:
         label = paddle.cast(label, paddle.int64)

python/paddle/nn/functional/loss.py

@@ -61,7 +61,7 @@ def dice_loss(input, label, epsilon=0.00001, name=None):
                              For more information, please refer to :ref:`api_guide_Name`
 
     Returns:
-        Tensor, which shape is [1], data type is the same as `input` .
+        0-D Tensor, which shape is [], data type is the same as `input` .
 
     Example:
         .. code-block:: python
@@ -363,7 +363,7 @@ def npair_loss(anchor, positive, labels, l2_reg=0.002):
 
 
     Returns:
-      A Tensor representing the npair loss, the data type is the same as anchor, the shape is [1].
+      A 0-D Tensor representing the npair loss, the data type is the same as anchor, the shape is [].
 
     Examples:
 
@@ -670,7 +670,7 @@ def binary_cross_entropy(
             input = paddle.to_tensor([0.5, 0.6, 0.7], 'float32')
             label = paddle.to_tensor([1.0, 0.0, 1.0], 'float32')
             output = paddle.nn.functional.binary_cross_entropy(input, label)
-            print(output)  # [0.65537095]
+            print(output)  # 0.65537095
 
     """
     if reduction not in ['sum', 'mean', 'none']:
@@ -810,7 +810,7 @@ def binary_cross_entropy_with_logits(
             logit = paddle.to_tensor([5.0, 1.0, 3.0])
             label = paddle.to_tensor([1.0, 0.0, 1.0])
             output = paddle.nn.functional.binary_cross_entropy_with_logits(logit, label)
-            print(output)  # [0.45618808]
+            print(output)  # 0.45618808
 
     """
     if reduction not in ['sum', 'mean', 'none']:
@@ -1113,7 +1113,7 @@ def smooth_l1_loss(input, label, reduction='mean', delta=1.0, name=None):
             label = paddle.rand([3, 3]).astype('float32')
             output = paddle.nn.functional.smooth_l1_loss(input, label)
             print(output)
-            # [0.068004]
+            # 0.068004
     """
 
     if in_dygraph_mode():
@@ -1183,7 +1183,7 @@ def margin_ranking_loss(
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
 
     Returns:
-        Tensor, if :attr:`reduction` is ``'mean'`` or ``'sum'``, the out shape is :math:`[1]`, otherwise the shape is the same as `input` .The same dtype as input tensor.
+        Tensor, if :attr:`reduction` is ``'mean'`` or ``'sum'``, the out shape is :math:`[]`, otherwise the shape is the same as `input` .The same dtype as input tensor.
 
     Examples:
 
@@ -1195,7 +1195,7 @@ def margin_ranking_loss(
             other = paddle.to_tensor([[2, 1], [2, 4]], dtype='float32')
             label = paddle.to_tensor([[1, -1], [-1, -1]], dtype='float32')
             loss = paddle.nn.functional.margin_ranking_loss(input, other, label)
-            print(loss) # [0.75]
+            print(loss) # 0.75
     """
     if reduction not in ['sum', 'mean', 'none']:
         raise ValueError(
@@ -1300,7 +1300,7 @@ def l1_loss(input, label, reduction='mean', name=None):
     Returns:
         Tensor, the L1 Loss of Tensor ``input`` and ``label``.
         If `reduction` is ``'none'``, the shape of output loss is :math:`[N, *]`, the same as ``input`` .
-        If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [1].
+        If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [].
 
     Examples:
         .. code-block:: python
@@ -1312,8 +1312,8 @@ def l1_loss(input, label, reduction='mean', name=None):
 
             l1_loss = paddle.nn.functional.l1_loss(input, label)
             print(l1_loss)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.34999999])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.34999999)
 
             l1_loss = paddle.nn.functional.l1_loss(input, label, reduction='none')
             print(l1_loss)
@@ -1323,8 +1323,8 @@ def l1_loss(input, label, reduction='mean', name=None):
 
             l1_loss = paddle.nn.functional.l1_loss(input, label, reduction='sum')
             print(l1_loss)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.39999998])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.39999998)
 
     """
     if reduction not in ['sum', 'mean', 'none']:
@@ -1413,7 +1413,7 @@ def nll_loss(
                 log_out = log_softmax(input)
                 label = paddle.to_tensor([0, 2, 1, 1, 0], "int64")
                 result = nll_loss(log_out, label)
-                print(result) # Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True, [1.07202101])
+                print(result) # Tensor(shape=[], dtype=float32, place=CPUPlace, stop_gradient=True, 1.07202101)
     """
     if reduction not in ['sum', 'mean', 'none']:
         raise ValueError(
@@ -1614,9 +1614,9 @@ def kl_div(input, label, reduction='mean', name=None):
 
     If `reduction` is ``'none'``, the output loss is the same shape as the input, and the loss at each point is calculated separately. There is no reduction to the result.
 
-    If `reduction` is ``'mean'``, the output loss is the shape of [1], and the output is the average of all losses.
+    If `reduction` is ``'mean'``, the output loss is the shape of [], and the output is the average of all losses.
 
-    If `reduction` is ``'sum'``, the output loss is the shape of [1], and the output is the sum of all losses.
+    If `reduction` is ``'sum'``, the output loss is the shape of [], and the output is the sum of all losses.
 
     If `reduction` is ``'batchmean'``, the output loss is the shape of [N], N is the batch size, and the output is the sum of all losses divided by the batch size.
 
@@ -1647,17 +1647,17 @@ def kl_div(input, label, reduction='mean', name=None):
             x = paddle.uniform(shape, min=-10, max=10).astype('float32')
             target = paddle.uniform(shape, min=-10, max=10).astype('float32')
 
-            # 'batchmean' reduction, loss shape will be [1]
+            # 'batchmean' reduction, loss shape will be [], who is 0-D Tensor
             pred_loss = F.kl_div(x, target, reduction='batchmean')
-            # shape=[1]
+            # shape=[]
 
-            # 'mean' reduction, loss shape will be [1]
+            # 'mean' reduction, loss shape will be [], who is 0-D Tensor
             pred_loss = F.kl_div(x, target, reduction='mean')
-            # shape=[1]
+            # shape=[]
 
-            # 'sum' reduction, loss shape will be [1]
+            # 'sum' reduction, loss shape will be [], who is 0-D Tensor
             pred_loss = F.kl_div(x, target, reduction='sum')
-            # shape=[1]
+            # shape=[]
 
             # 'none' reduction, loss shape is same with input shape
             pred_loss = F.kl_div(x, target, reduction='none')
@@ -1760,7 +1760,7 @@ def mse_loss(input, label, reduction='mean', name=None):
             label = paddle.to_tensor(1.7)
             output = mse_loss(input, label)
             print(output)
-            # [0.04000002]
+            # 0.04000002
 
     """
 
@@ -1816,7 +1816,7 @@ def ctc_loss(
         norm_by_times (bool, optional): Whether to normalize the gradients by the number of time-step, which is also the sequence's length. There is no need to normalize the gradients if reduction mode is 'mean'. Default: False.
 
     Returns:
-        Tensor, The Connectionist Temporal Classification (CTC) loss between ``log_probs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is [1]. Data type is the same as ``log_probs``.
+        Tensor, The Connectionist Temporal Classification (CTC) loss between ``log_probs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is []. Data type is the same as ``log_probs``.
 
     Examples:
 
@@ -1870,8 +1870,8 @@ def ctc_loss(
                 blank=0,
                 reduction='mean')
             print(loss)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.13760614])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.13760614)
 
     """
 
@@ -1965,7 +1965,7 @@ def rnnt_loss(
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
 
     Returns:
-        Tensor, The RNN-T loss between ``logprobs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is [1]. Data type is the same as ``logprobs``.
+        Tensor, The RNN-T loss between ``logprobs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is []. Data type is the same as ``logprobs``.
 
     Examples:
 
@@ -1997,8 +1997,8 @@ def rnnt_loss(
 
             costs = fn(acts, labels, lengths, label_lengths)
             print(costs)
-            # Tensor(shape=[1], dtype=float64, place=Place(gpu:0), stop_gradient=False,
-            #        [4.49566677])
+            # Tensor(shape=[], dtype=float64, place=Place(gpu:0), stop_gradient=False,
+            #        4.49566677)
     """
 
     def warprnnt(
@@ -2114,7 +2114,7 @@ def margin_cross_entropy(
             softmax is shard_softmax when using model parallel, otherwise
             softmax is in the same shape with input logits. If
             ``reduction == None``, the shape of loss is ``[N, 1]``, otherwise
-            the shape is ``[1]``.
+            the shape is ``[]``.
 
     Examples:
 
@@ -2669,8 +2669,8 @@ def cross_entropy(
                                         input,
                                         label)
             print(dy_ret)
-            # Tensor(shape=[1], dtype=float64, place=Place(gpu:0), stop_gradient=True,
-            #        [5.34043430])
+            # Tensor(shape=[], dtype=float64, place=Place(gpu:0), stop_gradient=True,
+            #        5.34043430)
 
         .. code-block:: python
 
@@ -2695,8 +2695,8 @@ def cross_entropy(
                                                                     weight=weight,
                                                                     reduction=reduction)
             print(paddle_loss_mean)
-            # Tensor(shape=[1], dtype=float64, place=Place(gpu:0), stop_gradient=True,
-            #        [1.11043464])
+            # Tensor(shape=[], dtype=float64, place=Place(gpu:0), stop_gradient=True,
+            #        1.11043464)
 
     """
 
@@ -3082,7 +3082,7 @@ def sigmoid_focal_loss(
             For more information, please refer to :ref:`api_guide_Name`.
 
     Returns:
-        Tensor, if :attr:`reduction` is ``'mean'`` or ``'sum'``, the out shape is :math:`[1]`, otherwise the shape is the same as ``logit``. The same dtype as ``logit`` tensor.
+        Tensor, if :attr:`reduction` is ``'mean'`` or ``'sum'``, the out shape is :math:`[]`, otherwise the shape is the same as ``logit``. The same dtype as ``logit`` tensor.
 
     Examples:
 
@@ -3096,7 +3096,7 @@ def sigmoid_focal_loss(
             fg_label = paddle.greater_equal(label, one)
             fg_num = paddle.sum(paddle.cast(fg_label, dtype='float32'))
             output = paddle.nn.functional.sigmoid_focal_loss(logit, label, normalizer=fg_num)
-            print(output)  # [0.65782464]
+            print(output)  # 0.65782464
 
     """
     if reduction not in ['sum', 'mean', 'none']:
@@ -3253,7 +3253,7 @@ def multi_label_soft_margin_loss(
             # Tensor([3.49625897, 0.71111226, 0.43989015])
             loss = F.multi_label_soft_margin_loss(input, label, reduction='mean')
             print(loss)
-            # Tensor([1.54908717])
+            # Tensor(1.54908717)
     """
     if reduction not in ['sum', 'mean', 'none']:
         raise ValueError(
@@ -3377,7 +3377,7 @@ def hinge_embedding_loss(input, label, margin=1.0, reduction='mean', name=None):
 
             loss = F.hinge_embedding_loss(input, label, margin=1.0, reduction='mean')
             print(loss)
-            # Tensor([0.22222222])
+            # Tensor(0.22222222)
     """
 
     if reduction not in ['sum', 'mean', 'none']:
@@ -3447,7 +3447,7 @@ def cosine_embedding_loss(
     Returns:
         Tensor, the cosine embedding Loss of Tensor ``input1`` ``input2`` and ``label``.
             If `reduction` is ``'none'``, the shape of output loss is [N], the same as ``input`` .
-            If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [1].
+            If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [].
 
     Examples:
         .. code-block:: python
@@ -3459,10 +3459,10 @@ def cosine_embedding_loss(
             label = paddle.to_tensor([1, -1], 'int64')
 
             output = paddle.nn.functional.cosine_embedding_loss(input1, input2, label, margin=0.5, reduction='mean')
-            print(output)  # [0.21155193]
+            print(output)  # 0.21155193
 
             output = paddle.nn.functional.cosine_embedding_loss(input1, input2, label, margin=0.5, reduction='sum')
-            print(output)  # [0.42310387]
+            print(output)  # 0.42310387
 
             output = paddle.nn.functional.cosine_embedding_loss(input1, input2, label, margin=0.5, reduction='none')
             print(output)  # [0.42310387, 0.        ]
@@ -3598,7 +3598,7 @@ def triplet_margin_with_distance_loss(
 
             loss = F.triplet_margin_with_distance_loss(input, positive, negative, margin=1.0, reduction='mean')
             print(loss)
-            # Tensor([0.19165580])
+            # Tensor(0.19165580)
 
     """
     if reduction not in ['sum', 'mean', 'none']:
@@ -3748,7 +3748,7 @@ def triplet_margin_loss(
 
             loss = F.triplet_margin_loss(input, positive, negative, margin=1.0, reduction='mean')
             print(loss)
-            # Tensor([0.19165580])
+            # Tensor(0.19165580)
 
     """
     if reduction not in ['sum', 'mean', 'none']:
@@ -3956,7 +3956,7 @@ def soft_margin_loss(input, label, reduction='mean', name=None):
 
     Returns:
 
-        Output (Tensor): If ``reduction`` is ``'none'``, the shape of output is same as ``input`` , else the shape of output is [1].
+        Output (Tensor): If ``reduction`` is ``'none'``, the shape of output is same as ``input`` , else the shape of output is [].
 
     Examples:
         .. code-block:: python
@@ -3967,8 +3967,8 @@ def soft_margin_loss(input, label, reduction='mean', name=None):
             label = paddle.to_tensor([[1.0, -1.0, 1.0],[-1.0, 1.0, 1.0]], 'float32')
             output = paddle.nn.functional.soft_margin_loss(input, label)
             print(output)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.64022040])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.64022040)
 
             input = paddle.uniform(shape=(5, 5), dtype="float32", min=0.1, max=0.8)
             label = paddle.randint(0, 2, shape=(5, 5), dtype="int64")
@@ -4068,7 +4068,7 @@ def gaussian_nll_loss(
 
     Returns:
 
-        output (Tensor): If ``reduction`` is ``'none'``, the shape of output is same as ``input`` , else the shape of output is [1].
+        output (Tensor): If ``reduction`` is ``'none'``, the shape of output is same as ``input`` , else the shape of output is [].
 
     Examples::
         .. code-block:: python

python/paddle/nn/layer/loss.py

@@ -98,8 +98,8 @@ class BCEWithLogitsLoss(Layer):
             bce_logit_loss = paddle.nn.BCEWithLogitsLoss()
             output = bce_logit_loss(logit, label)
             print(output)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.45618814])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.45618814)
 
     """
 
@@ -319,8 +319,8 @@ class CrossEntropyLoss(Layer):
                                         input,
                                         label)
             print(dy_ret)
-            # Tensor(shape=[1], dtype=float64, place=Place(gpu:0), stop_gradient=True,
-            #        [5.34043430])
+            # Tensor(shape=[], dtype=float64, place=Place(gpu:0), stop_gradient=True,
+            #        5.34043430)
 
         .. code-block:: python
 
@@ -345,8 +345,8 @@ class CrossEntropyLoss(Layer):
                                                                     weight=weight,
                                                                     reduction=reduction)
             print(paddle_loss_mean)
-            # Tensor(shape=[1], dtype=float64, place=Place(gpu:0), stop_gradient=True,
-            #        [1.11043464])
+            # Tensor(shape=[], dtype=float64, place=Place(gpu:0), stop_gradient=True,
+            #        1.11043464)
 
     """
 
@@ -564,7 +564,7 @@ class MSELoss(Layer):
             label = paddle.to_tensor([1.7])
             output = mse_loss(input, label)
             print(output)
-            # [0.04000002]
+            # 0.04000002
 
     """
 
@@ -637,7 +637,7 @@ class L1Loss(Layer):
         - label (Tensor): label. The shapes is ``[N, *]``, same shape as ``input`` . It's data type should be float32, float64, int32, int64.
         - output (Tensor): The L1 Loss of ``input`` and ``label``.
           If `reduction` is ``'none'``, the shape of output loss is ``[N, *]``, the same as ``input`` .
-          If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [1].
+          If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [].
 
     Examples:
         .. code-block:: python
@@ -650,14 +650,14 @@ class L1Loss(Layer):
             l1_loss = paddle.nn.L1Loss()
             output = l1_loss(input, label)
             print(output)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.34999999])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.34999999)
 
             l1_loss = paddle.nn.L1Loss(reduction='sum')
             output = l1_loss(input, label)
             print(output)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.39999998])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.39999998)
 
             l1_loss = paddle.nn.L1Loss(reduction='none')
             output = l1_loss(input, label)
@@ -747,8 +747,8 @@ class BCELoss(Layer):
             bce_loss = paddle.nn.BCELoss()
             output = bce_loss(input, label)
             print(output)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.65537101])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.65537101)
 
     """
 
@@ -835,7 +835,7 @@ class NLLLoss(Layer):
             The data type is int64.
         - output (Tensor): the `negative log likelihood loss` between input `x` and `label`.
             If `reduction` is `'none'`, the shape is `[N, *]`.
-            If `reduction` is `'sum'` or `'mean'`, the shape is `[1]`.
+            If `reduction` is `'sum'` or `'mean'`, the shape is `[]`.
 
     Examples:
         .. code-block:: python
@@ -853,7 +853,7 @@ class NLLLoss(Layer):
                 log_out = log_softmax(input)
                 label = paddle.to_tensor([0, 2, 1, 1, 0], "int64")
                 result = nll_loss(log_out, label)
-                print(result) # Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True, [1.07202101])
+                print(result) # Tensor(shape=[], dtype=float32, place=CPUPlace, stop_gradient=True, 1.07202101)
 
     """
 
@@ -991,9 +991,9 @@ class KLDivLoss(Layer):
 
     If `reduction` is ``'none'``, the output loss is the same shape as the input, and the loss at each point is calculated separately. There is no reduction to the result.
 
-    If `reduction` is ``'mean'``, the output loss is the shape of [1], and the output is the average of all losses.
+    If `reduction` is ``'mean'``, the output loss is the shape of [], and the output is the average of all losses.
 
-    If `reduction` is ``'sum'``, the output loss is the shape of [1], and the output is the sum of all losses.
+    If `reduction` is ``'sum'``, the output loss is the shape of [], and the output is the sum of all losses.
 
     If `reduction` is ``'batchmean'``, the output loss is the shape of [N], N is the batch size, and the output is the sum of all losses divided by the batch size.
 
@@ -1012,7 +1012,7 @@ class KLDivLoss(Layer):
 
         label (Tensor): ``(N, *)``, same shape as input.
 
-        output (Tensor): tensor with shape: [1] by default.
+        output (Tensor): tensor with shape: [] by default.
 
     Examples:
         .. code-block:: python
@@ -1024,20 +1024,20 @@ class KLDivLoss(Layer):
             x = paddle.uniform(shape, min=-10, max=10).astype('float32')
             target = paddle.uniform(shape, min=-10, max=10).astype('float32')
 
-            # 'batchmean' reduction, loss shape will be [1]
+            # 'batchmean' reduction, loss shape will be []
             kldiv_criterion = nn.KLDivLoss(reduction='batchmean')
             pred_loss = kldiv_criterion(x, target)
-            # shape=[1]
+            # shape=[]
 
-            # 'mean' reduction, loss shape will be [1]
+            # 'mean' reduction, loss shape will be []
             kldiv_criterion = nn.KLDivLoss(reduction='mean')
             pred_loss = kldiv_criterion(x, target)
-            # shape=[1]
+            # shape=[]
 
-            # 'sum' reduction, loss shape will be [1]
+            # 'sum' reduction, loss shape will be []
             kldiv_criterion = nn.KLDivLoss(reduction='sum')
             pred_loss = kldiv_criterion(x, target)
-            # shape=[1]
+            # shape=[]
 
             # 'none' reduction, loss shape is same with X shape
             kldiv_criterion = nn.KLDivLoss(reduction='none')
@@ -1090,7 +1090,7 @@ class MarginRankingLoss(Layer):
 
         label: N-D Tensor, label have the same shape and dtype as `input`.
 
-        output: If :attr:`reduction` is ``'mean'`` or ``'sum'`` , the out shape is :math:`[1]`, otherwise the shape is the same as `input` .The same dtype as input tensor.
+        output: If :attr:`reduction` is ``'mean'`` or ``'sum'`` , the out shape is :math:`[]`, otherwise the shape is the same as `input` .The same dtype as input tensor.
 
     Returns:
         A callable object of MarginRankingLoss.
@@ -1108,7 +1108,7 @@ class MarginRankingLoss(Layer):
             loss = margin_rank_loss(input, other, label)
 
             print(loss)
-            # [0.75]
+            # 0.75
     """
 
     def __init__(self, margin=0.0, reduction='mean', name=None):
@@ -1149,7 +1149,7 @@ class CTCLoss(Layer):
         - norm_by_times (bool, optional): Whether to normalize the gradients by the number of time-step, which is also the sequence's length. There is no need to normalize the gradients if reduction mode is 'mean'. Default: False.
 
     Returns:
-        Tensor, The Connectionist Temporal Classification (CTC) loss between ``log_probs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is [1]. Data type is the same as ``log_probs``.
+        Tensor, The Connectionist Temporal Classification (CTC) loss between ``log_probs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is []. Data type is the same as ``log_probs``.
 
     Examples:
 
@@ -1197,8 +1197,8 @@ class CTCLoss(Layer):
                 input_lengths,
                 label_lengths)
             print(loss)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.13760614])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.13760614)
     """
 
     def __init__(self, blank=0, reduction='mean'):
@@ -1242,7 +1242,7 @@ class RNNTLoss(Layer):
         label_lengths: Tensor of (batch) containing label length of each example
 
     Returns:
-     Tensor, The RNN-T loss between ``logprobs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is [1]. Data type is the same as ``logprobs``.
+     Tensor, The RNN-T loss between ``logprobs`` and  ``labels``. If attr:`reduction` is ``'none'``, the shape of loss is [batch_size], otherwise, the shape of loss is []. Data type is the same as ``logprobs``.
 
     Examples:
         .. code-block:: python
@@ -1272,8 +1272,8 @@ class RNNTLoss(Layer):
 
             costs = fn(acts, labels, lengths, label_lengths)
             print(costs)
-            # Tensor(shape=[1], dtype=float64, place=Place(gpu:0), stop_gradient=False,
-            #        [4.49566677])
+            # Tensor(shape=[], dtype=float64, place=Place(gpu:0), stop_gradient=False,
+            #        4.49566677)
     """
 
     def __init__(
@@ -1352,7 +1352,7 @@ class SmoothL1Loss(Layer):
             loss = paddle.nn.SmoothL1Loss()
             output = loss(input, label)
             print(output)
-            # [0.049606]
+            # 0.049606
     """
 
     def __init__(self, reduction='mean', delta=1.0, name=None):
@@ -1428,7 +1428,7 @@ class MultiLabelSoftMarginLoss(Layer):
                 multi_label_soft_margin_loss = nn.MultiLabelSoftMarginLoss(reduction='mean')
                 loss = multi_label_soft_margin_loss(input, label)
                 print(loss)
-                # Tensor([1.54908717])
+                # Tensor(1.54908717)
         """
 
     def __init__(self, weight=None, reduction="mean", name=None):
@@ -1529,7 +1529,7 @@ class HingeEmbeddingLoss(Layer):
             hinge_embedding_loss = nn.HingeEmbeddingLoss(margin=1.0, reduction='mean')
             loss = hinge_embedding_loss(input, label)
             print(loss)
-            # Tensor([0.22222222])
+            # Tensor(0.22222222)
     """
 
     def __init__(self, margin=1.0, reduction="mean", name=None):
@@ -1590,7 +1590,7 @@ class CosineEmbeddingLoss(Layer):
                          Available dtypes are int32, int64, float32, float64.
         output (Tensor): Tensor, the cosine embedding Loss of Tensor ``input1`` ``input2`` and ``label``.
                          If `reduction` is ``'none'``, the shape of output loss is [N], the same as ``input`` .
-                         If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [1].
+                         If `reduction` is ``'mean'`` or ``'sum'``, the shape of output loss is [].
 
     Examples:
         .. code-block:: python
@@ -1603,11 +1603,11 @@ class CosineEmbeddingLoss(Layer):
 
             cosine_embedding_loss = paddle.nn.CosineEmbeddingLoss(margin=0.5, reduction='mean')
             output = cosine_embedding_loss(input1, input2, label)
-            print(output) # [0.21155193]
+            print(output) # 0.21155193
 
             cosine_embedding_loss = paddle.nn.CosineEmbeddingLoss(margin=0.5, reduction='sum')
             output = cosine_embedding_loss(input1, input2, label)
-            print(output) # [0.42310387]
+            print(output) # 0.42310387
 
             cosine_embedding_loss = paddle.nn.CosineEmbeddingLoss(margin=0.5, reduction='none')
             output = cosine_embedding_loss(input1, input2, label)
@@ -1717,7 +1717,7 @@ class TripletMarginWithDistanceLoss(Layer):
             triplet_margin_with_distance_loss = TripletMarginWithDistanceLoss(reduction='mean')
             loss = triplet_margin_with_distance_loss(input, positive, negative,)
             print(loss)
-            # Tensor([0.19165580])
+            # Tensor(0.19165580)
 
     """
 
@@ -1825,7 +1825,7 @@ class TripletMarginLoss(Layer):
             triplet_margin_loss = paddle.nn.TripletMarginLoss(margin=1.0, swap=True, reduction='mean', )
             loss = triplet_margin_loss(input, positive, negative,)
             print(loss)
-            # Tensor([0.19165580])
+            # Tensor(0.19165580)
 
     """
 
@@ -1995,7 +1995,7 @@ class SoftMarginLoss(Layer):
           ``input``. The target labels which values should be numbers -1 or 1.
           Available dtype is int32, int64, float32, float64.
         - Output (Tensor): If ``reduction`` is ``'none'``, the shape of output is
-          same as ``input`` , else the shape of output is [1].
+          same as ``input`` , else the shape of output is [].
 
     Returns:
         A callable object of SoftMarginLoss.
@@ -2010,8 +2010,8 @@ class SoftMarginLoss(Layer):
             soft_margin_loss = paddle.nn.SoftMarginLoss()
             output = soft_margin_loss(input, label)
             print(output)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.64022040])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.64022040)
 
             input_np = paddle.uniform(shape=(5, 5), min=0.1, max=0.8, dtype="float64")
             label_np = paddle.randint(high=2, shape=(5, 5), dtype="int64")

python/paddle/static/nn/metric.py

@@ -66,10 +66,10 @@ def accuracy(input, label, k=1, correct=None, total=None):
             exe.run(static.default_startup_program())
             x = np.random.rand(3, 32, 32).astype("float32")
             y = np.array([[1],[0],[1]])
-            output= exe.run(feed={"input": x,"label": y},
-                        fetch_list=[result[0]])
+            output = exe.run(feed={"input": x,"label": y},
+                             fetch_list=[result])
             print(output)
-            #[array([0.], dtype=float32)]
+            # [array(0.33333334, dtype=float32)]
 
     """
     if _non_static_mode():

python/paddle/tensor/creation.py

@@ -711,10 +711,10 @@ def to_tensor(data, dtype=None, place=None, stop_gradient=True):
         We use the dtype conversion rules following this:
                 Keep dtype
         np.number ───────────► paddle.Tensor
-                                (0D-Tensor)
+                                (0-D Tensor)
                     default_dtype
         Python Number ───────────────► paddle.Tensor
-                                        (0D-Tensor)
+                                        (0-D Tensor)
                     Keep dtype
         np.ndarray ───────────► paddle.Tensor
 
@@ -747,7 +747,6 @@ def to_tensor(data, dtype=None, place=None, stop_gradient=True):
         #        1)
 
         x = paddle.to_tensor(1, stop_gradient=False)
-        print(x)
         # Tensor(shape=[], dtype=int64, place=CPUPlace, stop_gradient=False,
         #        1)
 

python/paddle/tensor/linalg.py

@@ -335,8 +335,8 @@ def norm(x, p='fro', axis=None, keepdim=False, name=None):
 
             # compute inf-order  norm
             out_pnorm = paddle.linalg.norm(x, p=float("inf"))
-            # out_pnorm  = Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #                    [12.])
+            # out_pnorm  = Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #                    12.)
 
             out_pnorm = paddle.linalg.norm(x, p=float("inf"), axis=0)
             # out_pnorm: Tensor(shape=[3, 4], dtype=float32, place=Place(cpu), stop_gradient=True,
@@ -346,8 +346,8 @@ def norm(x, p='fro', axis=None, keepdim=False, name=None):
 
             # compute -inf-order  norm
             out_pnorm = paddle.linalg.norm(x, p=-float("inf"))
-            # out_pnorm: Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #                  [0.])
+            # out_pnorm: Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #                  0.)
 
             out_pnorm = paddle.linalg.norm(x, p=-float("inf"), axis=0)
             # out_pnorm: Tensor(shape=[3, 4], dtype=float32, place=Place(cpu), stop_gradient=True,
@@ -692,16 +692,16 @@ def dist(x, y, p=2, name=None):
             x = paddle.to_tensor([[3, 3],[3, 3]], dtype="float32")
             y = paddle.to_tensor([[3, 3],[3, 1]], dtype="float32")
             out = paddle.dist(x, y, 0)
-            print(out) # out = [1.]
+            print(out) # out = 1.
 
             out = paddle.dist(x, y, 2)
-            print(out) # out = [2.]
+            print(out) # out = 2.
 
             out = paddle.dist(x, y, float("inf"))
-            print(out) # out = [2.]
+            print(out) # out = 2.
 
             out = paddle.dist(x, y, float("-inf"))
-            print(out) # out = [0.]
+            print(out) # out = 0.
     """
     if in_dygraph_mode():
         return _C_ops.dist(x, y, p)
@@ -747,48 +747,48 @@ def cond(x, p=None, name=None):
 
             # compute conditional number when p is None
             out = paddle.linalg.cond(x)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.41421342])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.41421342)
 
             # compute conditional number when order of the norm is 'fro'
             out_fro = paddle.linalg.cond(x, p='fro')
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [3.16227770])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        3.16227770)
 
             # compute conditional number when order of the norm is 'nuc'
             out_nuc = paddle.linalg.cond(x, p='nuc')
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [9.24263859])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        9.24263859)
 
             # compute conditional number when order of the norm is 1
             out_1 = paddle.linalg.cond(x, p=1)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [2.])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        2.)
 
             # compute conditional number when order of the norm is -1
             out_minus_1 = paddle.linalg.cond(x, p=-1)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.)
 
             # compute conditional number when order of the norm is 2
             out_2 = paddle.linalg.cond(x, p=2)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.41421342])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.41421342)
 
             # compute conditional number when order of the norm is -1
             out_minus_2 = paddle.linalg.cond(x, p=-2)
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [0.70710683])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        0.70710683)
 
             # compute conditional number when order of the norm is inf
             out_inf = paddle.linalg.cond(x, p=float("inf"))
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [2.])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        2.)
 
             # compute conditional number when order of the norm is -inf
             out_minus_inf = paddle.linalg.cond(x, p=-float("inf"))
-            # Tensor(shape=[1], dtype=float32, place=Place(gpu:0), stop_gradient=True,
-            #        [1.])
+            # Tensor(shape=[], dtype=float32, place=Place(gpu:0), stop_gradient=True,
+            #        1.)
 
             a = paddle.randn([2, 4, 4])
             # Tensor(shape=[2, 4, 4], dtype=float32, place=Place(gpu:0), stop_gradient=True,
@@ -1097,13 +1097,13 @@ def dot(x, y, name=None):
         x = paddle.to_tensor([1, 2, 3])
         y = paddle.to_tensor([4, 5, 6])
         z = paddle.dot(x, y)
-        print(z)  # [32]
+        print(z)  # 32
 
         # 2-D Tensor * 2-D Tensor
         x = paddle.to_tensor([[1, 2, 3], [2, 4, 6]])
         y = paddle.to_tensor([[4, 5, 6], [4, 5, 6]])
         z = paddle.dot(x, y)
-        print(z)  # [[32], [64]]
+        print(z)  # [32, 64]
 
     """
     if in_dygraph_mode():
@@ -1165,7 +1165,7 @@ def cov(x, rowvar=True, ddof=True, fweights=None, aweights=None, name=None):
 
         import paddle
 
-        xt = paddle.rand((3,4))
+        xt = paddle.rand((3, 4))
         paddle.linalg.cov(xt)
 
         '''
@@ -1487,7 +1487,7 @@ def matrix_rank(x, tol=None, hermitian=False, name=None):
             a = paddle.eye(10)
             b = paddle.linalg.matrix_rank(a)
             print(b)
-            # b = [10]
+            # b = 10
 
             c = paddle.ones(shape=[3, 4, 5, 5])
             d = paddle.linalg.matrix_rank(c, tol=0.01, hermitian=True)

python/paddle/tensor/logic.py

@@ -288,13 +288,8 @@ def is_empty(x, name=None):
 
             input = paddle.rand(shape=[4, 32, 32], dtype='float32')
             res = paddle.is_empty(x=input)
-            print("res:", res)
-            # ('res:', Tensor: eager_tmp_1
-            #    - place: CPUPlace
-            #    - shape: [1]
-            #    - layout: NCHW
-            #    - dtype: bool
-            #    - data: [0])
+            # res: Tensor(shape=[], dtype=bool, place=Place(cpu), stop_gradient=True,
+            #        False)
 
     """
     if in_dygraph_mode():
@@ -339,9 +334,9 @@ def equal_all(x, y, name=None):
           y = paddle.to_tensor([1, 2, 3])
           z = paddle.to_tensor([1, 4, 3])
           result1 = paddle.equal_all(x, y)
-          print(result1) # result1 = [True ]
+          print(result1) # result1 = True
           result2 = paddle.equal_all(x, z)
-          print(result2) # result2 = [False ]
+          print(result2) # result2 = False
     """
     if in_dygraph_mode():
         return _C_ops.equal_all(x, y)
@@ -388,21 +383,21 @@ def allclose(x, y, rtol=1e-05, atol=1e-08, equal_nan=False, name=None):
           y = paddle.to_tensor([10000.1, 1e-08])
           result1 = paddle.allclose(x, y, rtol=1e-05, atol=1e-08,
                                   equal_nan=False, name="ignore_nan")
-          # [False]
+          # False
 
           result2 = paddle.allclose(x, y, rtol=1e-05, atol=1e-08,
                                       equal_nan=True, name="equal_nan")
-          # [False]
+          # False
 
           x = paddle.to_tensor([1.0, float('nan')])
           y = paddle.to_tensor([1.0, float('nan')])
           result1 = paddle.allclose(x, y, rtol=1e-05, atol=1e-08,
                                   equal_nan=False, name="ignore_nan")
-          # [False]
+          # False
 
           result2 = paddle.allclose(x, y, rtol=1e-05, atol=1e-08,
                                       equal_nan=True, name="equal_nan")
-          # [True]
+          # True
     """
 
     if in_dygraph_mode():

python/paddle/tensor/manipulation.py

@@ -268,11 +268,11 @@ def slice(input, axes, starts, ends):
     Args:
         input (Tensor): A ``Tensor`` . The data type is ``float16``, ``float32``, ``float64``, ``int32`` or ``int64``.
         axes (list|tuple): The data type is ``int32`` . Axes that `starts` and `ends` apply to .
-        starts (list|tuple|Tensor): The data type is ``int32`` . If ``starts`` is a list or tuple, the elements of
-                it should be integers or Tensors with shape [1]. If ``starts`` is an Tensor, it should be an 1-D Tensor.
+        starts (list|tuple|Tensor): The data type is ``int32`` . If ``starts`` is a list or tuple, each element of
+                it should be integer or 0-D int Tensor with shape []. If ``starts`` is an Tensor, it should be an 1-D Tensor.
                 It represents starting indices of corresponding axis in ``axes``.
-        ends (list|tuple|Tensor): The data type is ``int32`` . If ``ends`` is a list or tuple, the elements of
-                it should be integers or Tensors with shape [1]. If ``ends`` is an Tensor, it should be an 1-D Tensor .
+        ends (list|tuple|Tensor): The data type is ``int32`` . If ``ends`` is a list or tuple, each element of
+                it should be integer or 0-D int Tensor with shape []. If ``ends`` is an Tensor, it should be an 1-D Tensor .
                 It represents ending indices of corresponding axis in ``axes``.
 
     Returns:
@@ -1065,21 +1065,21 @@ def tolist(x):
             print(expectlist)   #[0, 1, 2, 3, 4]
 
     """
-    # TODO(zhouwei): will remove 0D Tensor.numpy() hack
+    # TODO(zhouwei): will remove 0-D Tensor.numpy() hack
     return x.numpy(False).tolist()
 
 
 def concat(x, axis=0, name=None):
     """
 
-    Concatenates the input along the axis.
+    Concatenates the input along the axis. It doesn't support 0-D Tensor because it requires a certain axis, and 0-D Tensor
+    doesn't have any axis.
 
     Args:
         x (list|tuple): ``x`` is a Tensor list or Tensor tuple which is with data type bool, float16,
             float32, float64, int32, int64, int8, uint8. All the Tensors in ``x`` must have same data type.
         axis (int|Tensor, optional): Specify the axis to operate on the input Tensors.
-            It's a scalar with data type int or a Tensor with shape [1] and data type int32
-            or int64. The effective range is [-R, R), where R is Rank(x). When ``axis < 0``,
+            Tt should be integer or 0-D int Tensor with shape []. The effective range is [-R, R), where R is Rank(x). When ``axis < 0``,
             it works the same way as ``axis+R``. Default is 0.
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
 
@@ -1550,11 +1550,11 @@ def flatten(x, start_axis=0, stop_axis=-1, name=None):
     if x_dim == 0:
         if not (isinstance(start_axis, int)) or start_axis not in [0, -1]:
             raise ValueError(
-                "The start_axis should be int, and should be 0 or -1 when the input tensor is a 0D-Tensor"
+                "The start_axis should be int, and should be 0 or -1 when the input tensor is a 0-D-Tensor"
             )
         if not (isinstance(stop_axis, int)) or stop_axis not in [0, -1]:
             raise ValueError(
-                "The stop_axis should be int, and should be 0 or -1 when the input tensor is a 0D-Tensor"
+                "The stop_axis should be int, and should be 0 or -1 when the input tensor is a 0-D-Tensor"
             )
     else:
         if (
@@ -1913,8 +1913,8 @@ def split(x, num_or_sections, axis=0, name=None):
             If ``num_or_sections`` is a list or tuple, the length of it indicates the number of
             sub-Tensors and the elements in it indicate the sizes of sub-Tensors'  dimension orderly.
             The length of the list must not  be larger than the ``x`` 's size of specified ``axis``.
-        axis (int|Tensor, optional): The axis along which to split, it can be a scalar with type
-            ``int`` or a ``Tensor`` with shape [1] and data type  ``int32`` or ``int64``.
+        axis (int|Tensor, optional): The axis along which to split, it can be a integer or a ``0-D Tensor``
+            with shape [] and data type  ``int32`` or ``int64``.
             If :math::`axis < 0`, the axis to split along is :math:`rank(x) + axis`. Default is 0.
         name (str, optional): The default value is None.  Normally there is no need for user to set this property.
             For more information, please refer to :ref:`api_guide_Name` .
@@ -2557,7 +2557,7 @@ def unsqueeze(x, axis, name=None):
     Args:
         x (Tensor): The input Tensor to be unsqueezed. Supported data type: float32, float64, bool, int8, int32, int64.
         axis (int|list|tuple|Tensor): Indicates the dimensions to be inserted. The data type is ``int32`` .
-                                    If ``axis`` is a list or tuple, the elements of it should be integers or Tensors with shape [1].
+                                    If ``axis`` is a list or tuple, each element of it should be integer or 0-D Tensor with shape [].
                                     If ``axis`` is a Tensor, it should be an 1-D Tensor .
                                     If ``axis`` is negative, ``axis = axis + ndim(x) + 1``.
         name (str|None): Name for this layer. Please refer to :ref:`api_guide_Name`, Default None.
@@ -3082,8 +3082,8 @@ def chunk(x, chunks, axis=0, name=None):
     Args:
         x (Tensor): A N-D Tensor. The data type is bool, float16, float32, float64, int32 or int64.
         chunks(int): The number of tensor to be split along the certain axis.
-        axis (int|Tensor, optional): The axis along which to split, it can be a scalar with type
-            ``int`` or a ``Tensor`` with shape [1] and data type  ``int32`` or ``int64``.
+        axis (int|Tensor, optional): The axis along which to split, it can be a integer or a ``0-D Tensor``
+            with shape [] and data type  ``int32`` or ``int64``.
             If :math::`axis < 0`, the axis to split along is :math:`rank(x) + axis`. Default is 0.
         name (str, optional): The default value is None.  Normally there is no need for user to set this property.
             For more information, please refer to :ref:`api_guide_Name` .
@@ -3522,7 +3522,7 @@ def reshape(x, shape, name=None):
     Args:
         x (Tensor): An N-D Tensor. The data type is ``float32``, ``float64``, ``int32``, ``int64`` or ``bool``
         shape (list|tuple|Tensor): Define the target shape. At most one dimension of the target shape can be -1.
-                        The data type is ``int32`` . If ``shape`` is a list or tuple, the elements of it should be integers or Tensors with shape [].
+                        The data type is ``int32`` . If ``shape`` is a list or tuple, each element of it should be integer or Tensor with shape [].
                         If ``shape`` is an Tensor, it should be an 1-D Tensor .
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
 
@@ -3842,11 +3842,15 @@ def strided_slice(x, axes, starts, ends, strides, name=None):
         x (Tensor): An N-D ``Tensor``. The data type is ``bool``, ``float16``, ``float32``, ``float64``, ``int32`` or ``int64``.
         axes (list|tuple): The data type is ``int32`` . Axes that `starts` and `ends` apply to.
                             It's optional. If it is not provides, it will be treated as :math:`[0,1,...,len(starts)-1]`.
-        starts (list|tuple|Tensor): The data type is ``int32`` . If ``starts`` is a list or tuple, the elements of                                                                                          it should be integers or Tensors with shape [1]. If ``starts`` is an Tensor, it should be an 1-D Tensor.                                                                                    It represents starting indices of corresponding axis in ``axes``.
-        ends (list|tuple|Tensor): The data type is ``int32`` . If ``ends`` is a list or tuple, the elements of
-                it should be integers or Tensors with shape [1]. If ``ends`` is an Tensor, it should be an 1-D Tensor .                                                                                     It represents ending indices of corresponding axis in ``axes``.
-        strides (list|tuple|Tensor): The data type is ``int32`` . If ``strides`` is a list or tuple, the elements of
-                it should be integers or Tensors with shape [1]. If ``strides`` is an Tensor, it should be an 1-D Tensor .                                                                                  It represents slice step of corresponding axis in ``axes``.
+        starts (list|tuple|Tensor): The data type is ``int32`` . If ``starts`` is a list or tuple, the elements of it should be
+            integers or Tensors with shape []. If ``starts`` is an Tensor, it should be an 1-D Tensor.
+            It represents starting indices of corresponding axis in ``axes``.
+        ends (list|tuple|Tensor): The data type is ``int32`` . If ``ends`` is a list or tuple, the elements of it should be
+            integers or Tensors with shape []. If ``ends`` is an Tensor, it should be an 1-D Tensor.
+            It represents ending indices of corresponding axis in ``axes``.
+        strides (list|tuple|Tensor): The data type is ``int32`` . If ``strides`` is a list or tuple, the elements of it should be
+            integers or Tensors with shape []. If ``strides`` is an Tensor, it should be an 1-D Tensor.
+            It represents slice step of corresponding axis in ``axes``.
         name(str, optional): The default value is None.  Normally there is no need for user to set this property.
                         For more information, please refer to :ref:`api_guide_Name` .
 
@@ -4073,7 +4077,7 @@ def tensordot(x, y, axes=2, name=None):
             y = paddle.arange(10, dtype=data_type)
             z1 = paddle.tensordot(x, y, axes=1)
             z2 = paddle.dot(x, y)
-            # z1 = z2 = [285.]
+            # z1 = z2 = 285.
 
 
             # For two 2-d tensor x and y, the case axes=1 is equivalent to matrix multiplication.

python/paddle/tensor/math.py

@@ -187,7 +187,7 @@ def scale(x, scale=1.0, bias=0.0, bias_after_scale=True, act=None, name=None):
 
     Args:
         x (Tensor): Input N-D Tensor of scale operator. Data type can be float32, float64, int8, int16, int32, int64, uint8.
-        scale (float|Tensor): The scale factor of the input, it should be a float number or a Tensor with shape [1] and data type as float32.
+        scale (float|Tensor): The scale factor of the input, it should be a float number or a 0-D Tensor with shape [] and data type as float32.
         bias (float): The bias to be put on the input.
         bias_after_scale (bool): Apply bias addition after or before scaling. It is useful for numeric stability in some circumstances.
         act (str, optional): Activation applied to the output such as tanh, softmax, sigmoid, relu.
@@ -1278,9 +1278,9 @@ def sum(x, axis=None, dtype=None, keepdim=False, name=None):
             # Each example is followed by the corresponding output tensor.
             x = paddle.to_tensor([[0.2, 0.3, 0.5, 0.9],
                                   [0.1, 0.2, 0.6, 0.7]])
-            out1 = paddle.sum(x)  # [3.5]
+            out1 = paddle.sum(x)          # 3.5
             out2 = paddle.sum(x, axis=0)  # [0.3, 0.5, 1.1, 1.6]
-            out3 = paddle.sum(x, axis=-1)  # [1.9, 1.6]
+            out3 = paddle.sum(x, axis=-1) # [1.9, 1.6]
             out4 = paddle.sum(x, axis=1, keepdim=True)  # [[1.9], [1.6]]
 
             # y is a Tensor with shape [2, 2, 2] and elements as below:
@@ -1298,7 +1298,7 @@ def sum(x, axis=None, dtype=None, keepdim=False, name=None):
             # Each example is followed by the corresponding output tensor.
             x = paddle.to_tensor([[True, True, True, True],
                                   [False, False, False, False]])
-            out7 = paddle.sum(x)  # [4]
+            out7 = paddle.sum(x)          # 4
             out8 = paddle.sum(x, axis=0)  # [1, 1, 1, 1]
             out9 = paddle.sum(x, axis=1)  # [4, 0]
     """
@@ -1434,9 +1434,9 @@ def nansum(x, axis=None, dtype=None, keepdim=False, name=None):
             # Each example is followed by the corresponding output tensor.
             x = paddle.to_tensor([[float('nan'), 0.3, 0.5, 0.9],
                             [0.1, 0.2, float('-nan'), 0.7]],dtype="float32")
-            out1 = paddle.nansum(x)  # [2.7]
+            out1 = paddle.nansum(x)          # 2.7
             out2 = paddle.nansum(x, axis=0)  # [0.1, 0.5, 0.5, 1.6]
-            out3 = paddle.nansum(x, axis=-1)  # [1.7, 1.0]
+            out3 = paddle.nansum(x, axis=-1) # [1.7, 1.0]
             out4 = paddle.nansum(x, axis=1, keepdim=True)  # [[1.7], [1.0]]
 
             # y is a Tensor with shape [2, 2, 2] and elements as below:
@@ -1494,7 +1494,7 @@ def nanmean(x, axis=None, keepdim=False, name=None):
             x = paddle.to_tensor([[float('nan'), 0.3, 0.5, 0.9],
                                   [0.1, 0.2, float('-nan'), 0.7]])
             out1 = paddle.nanmean(x)
-            # [0.44999996]
+            # 0.44999996
             out2 = paddle.nanmean(x, axis=0)
             # [0.1, 0.25, 0.5, 0.79999995]
             out3 = paddle.nanmean(x, axis=0, keepdim=True)
@@ -2204,7 +2204,7 @@ def logsumexp(x, axis=None, keepdim=False, name=None):
         import paddle
 
         x = paddle.to_tensor([[-1.5, 0., 2.], [3., 1.2, -2.4]])
-        out1 = paddle.logsumexp(x) # [3.4691226]
+        out1 = paddle.logsumexp(x)    # 3.4691226
         out2 = paddle.logsumexp(x, 1) # [2.15317821, 3.15684602]
 
     """
@@ -2316,7 +2316,7 @@ def max(x, axis=None, keepdim=False, name=None):
             result1 = paddle.max(x)
             result1.backward()
             print(result1, x.grad)
-            #[0.9], [[0., 0., 0., 1.], [0., 0., 0., 0.]]
+            # 0.9, [[0., 0., 0., 1.], [0., 0., 0., 0.]]
 
             x.clear_grad()
             result2 = paddle.max(x, axis=0)
@@ -2417,7 +2417,7 @@ def min(x, axis=None, keepdim=False, name=None):
             result1 = paddle.min(x)
             result1.backward()
             print(result1, x.grad)
-            #[0.1], [[0., 0., 0., 0.], [1., 0., 0., 0.]]
+            # 0.1, [[0., 0., 0., 0.], [1., 0., 0., 0.]]
 
             x.clear_grad()
             result2 = paddle.min(x, axis=0)
@@ -2521,13 +2521,13 @@ def amax(x, axis=None, keepdim=False, name=None):
             result1 = paddle.amax(x)
             result1.backward()
             print(result1, x.grad)
-            #[0.9], [[0., 0.2, 0.2, 0.2], [0.2, 0.2, 0., 0.]]
+            # 0.9, [[0., 0.2, 0.2, 0.2], [0.2, 0.2, 0., 0.]]
 
             x.clear_grad()
             result1_max = paddle.max(x)
             result1_max.backward()
             print(result1_max, x.grad)
-            #[0.9], [[0., 1.0, 1.0, 1.0], [1.0, 1.0, 0., 0.]]
+            # 0.9, [[0., 1.0, 1.0, 1.0], [1.0, 1.0, 0., 0.]]
 
             ###############################
 
@@ -2631,13 +2631,13 @@ def amin(x, axis=None, keepdim=False, name=None):
             result1 = paddle.amin(x)
             result1.backward()
             print(result1, x.grad)
-            #[0.1], [[0., 0.2, 0.2, 0.2], [0.2, 0.2, 0., 0.]]
+            # 0.1, [[0., 0.2, 0.2, 0.2], [0.2, 0.2, 0., 0.]]
 
             x.clear_grad()
             result1_min = paddle.min(x)
             result1_min.backward()
             print(result1_min, x.grad)
-            #[0.1], [[0., 1.0, 1.0, 1.0], [1.0, 1.0, 0., 0.]]
+            # 0.1, [[0., 1.0, 1.0, 1.0], [1.0, 1.0, 0., 0.]]
 
             ###############################
 
@@ -2848,10 +2848,10 @@ def clip(x, min=None, max=None, name=None):
 
     Args:
         x (Tensor): An N-D Tensor with data type float16, float32, float64, int32 or int64.
-        min (float|int|Tensor, optional): The lower bound with type ``float`` , ``int`` or a ``Tensor``
-            with shape [1] and type ``int32``, ``float16``, ``float32``, ``float64``.
-        max (float|int|Tensor, optional): The upper bound with type ``float``, ``int`` or a ``Tensor``
-            with shape [1] and type ``int32``, ``float16``, ``float32``, ``float64``.
+        min (float|int|Tensor, optional): The lower bound with type ``float`` , ``int`` or a ``0-D Tensor``
+            with shape [] and type ``int32``, ``float16``, ``float32``, ``float64``.
+        max (float|int|Tensor, optional): The upper bound with type ``float``, ``int`` or a ``0-D Tensor``
+            with shape [] and type ``int32``, ``float16``, ``float32``, ``float64``.
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
 
     Returns:
@@ -3005,7 +3005,7 @@ def trace(x, offset=0, axis1=0, axis2=1, name=None):
             case1 = paddle.randn([2, 3])
             case2 = paddle.randn([3, 10, 10])
             case3 = paddle.randn([3, 10, 5, 10])
-            data1 = paddle.trace(case1) # data1.shape = [1]
+            data1 = paddle.trace(case1) # data1.shape = []
             data2 = paddle.trace(case2, offset=1, axis1=1, axis2=2) # data2.shape = [3]
             data3 = paddle.trace(case3, offset=-3, axis1=1, axis2=-1) # data2.shape = [3, 5]
     """
@@ -3633,7 +3633,7 @@ def prod(x, axis=None, keepdim=False, dtype=None, name=None):
             x = paddle.to_tensor([[0.2, 0.3, 0.5, 0.9],
                                   [0.1, 0.2, 0.6, 0.7]])
             out1 = paddle.prod(x)
-            # [0.0002268]
+            # 0.0002268
 
             out2 = paddle.prod(x, -1)
             # [0.027  0.0084]
@@ -3839,8 +3839,8 @@ def all(x, axis=None, keepdim=False, name=None):
             print(x)
             x = paddle.cast(x, 'bool')
 
-            # out1 should be [False]
-            out1 = paddle.all(x)  # [False]
+            # out1 should be False
+            out1 = paddle.all(x)          # False
             print(out1)
 
             # out2 should be [True, False]
@@ -3848,7 +3848,7 @@ def all(x, axis=None, keepdim=False, name=None):
             print(out2)
 
             # keepdim=False, out3 should be [False, True], out.shape should be (2,)
-            out3 = paddle.all(x, axis=-1)  # [False, True]
+            out3 = paddle.all(x, axis=-1) # [False, True]
             print(out3)
 
             # keepdim=True, out4 should be [[False], [True]], out.shape should be (2,1)
@@ -3913,12 +3913,12 @@ def any(x, axis=None, keepdim=False, name=None):
             #    [[True, False]
             #     [True, True]]
 
-            # out1 should be [True]
-            out1 = paddle.any(x)  # [True]
+            # out1 should be True
+            out1 = paddle.any(x)           # True
             print(out1)
 
             # out2 should be [True, True]
-            out2 = paddle.any(x, axis=0)  # [True, True]
+            out2 = paddle.any(x, axis=0)   # [True, True]
             print(out2)
 
             # keepdim=False, out3 should be [True, True], out.shape should be (2,)
@@ -4422,8 +4422,8 @@ def rad2deg(x, name=None):
             x2 = paddle.to_tensor(math.pi/2)
             result2 = paddle.rad2deg(x2)
             print(result2)
-            # Tensor(shape=[1], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
-            #         [90.])
+            # Tensor(shape=[], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
+            #         90.)
 
             x3 = paddle.to_tensor(1)
             result3 = paddle.rad2deg(x3)
@@ -5323,27 +5323,27 @@ def trapezoid(y, x=None, dx=None, axis=-1, name=None):
             y = paddle.to_tensor([4, 5, 6], dtype='float32')
 
             print(paddle.trapezoid(y))
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [10.])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        10.)
 
             print(paddle.trapezoid(y, dx=2.))
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [20.])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        20.)
 
             y = paddle.to_tensor([4, 5, 6], dtype='float32')
             x = paddle.to_tensor([1, 2, 3], dtype='float32')
 
             print(paddle.trapezoid(y, x))
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [10.])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        10.)
 
 
             y = paddle.to_tensor([1, 2, 3], dtype='float64')
             x = paddle.to_tensor([8, 6, 4], dtype='float64')
 
             print(paddle.trapezoid(y, x))
-            # Tensor(shape=[1], dtype=float64, place=Place(cpu), stop_gradient=True,
-            #        [-8.])
+            # Tensor(shape=[], dtype=float64, place=Place(cpu), stop_gradient=True,
+            #        -8.)
             y = paddle.arange(6).reshape((2, 3)).astype('float32')
 
             print(paddle.trapezoid(y, axis=0))

python/paddle/tensor/stat.py

@@ -65,7 +65,7 @@ def mean(x, axis=None, keepdim=False, name=None):
                                    [17., 18., 19., 20.],
                                    [21., 22., 23., 24.]]])
             out1 = paddle.mean(x)
-            # [12.5]
+            # 12.5
             out2 = paddle.mean(x, axis=-1)
             # [[ 2.5  6.5 10.5]
             #  [14.5 18.5 22.5]]
@@ -140,7 +140,7 @@ def var(x, axis=None, unbiased=True, keepdim=False, name=None):
 
             x = paddle.to_tensor([[1.0, 2.0, 3.0], [1.0, 4.0, 5.0]])
             out1 = paddle.var(x)
-            # [2.66666667]
+            # 2.66666667
             out2 = paddle.var(x, axis=1)
             # [1.         4.33333333]
     """
@@ -205,9 +205,9 @@ def std(x, axis=None, unbiased=True, keepdim=False, name=None):
 
             x = paddle.to_tensor([[1.0, 2.0, 3.0], [1.0, 4.0, 5.0]])
             out1 = paddle.std(x)
-            # [1.63299316]
+            # 1.63299316
             out2 = paddle.std(x, unbiased=False)
-            # [1.49071205]
+            # 1.49071205
             out3 = paddle.std(x, axis=1)
             # [1.       2.081666]
 
@@ -222,8 +222,7 @@ def std(x, axis=None, unbiased=True, keepdim=False, name=None):
 
 def numel(x, name=None):
     """
-    Returns the number of elements for a tensor, which is a int64 Tensor with shape [1] in static graph mode
-    or a scalar value in imperative mode.
+    Returns the number of elements for a tensor, which is a 0-D int64 Tensor with shape [].
 
     Args:
         x (Tensor): The input Tensor, it's data type can be bool, float16, float32, float64, int32, int64.
@@ -231,7 +230,7 @@ def numel(x, name=None):
             For more information, please refer to :ref:`api_guide_Name`.
 
     Returns:
-        Tensor: The number of elements for the input Tensor.
+        Tensor: The number of elements for the input Tensor, whose shape is [].
 
     Examples:
         .. code-block:: python
@@ -387,8 +386,8 @@ def median(x, axis=None, keepdim=False, name=None):
             #         [8 , 9 , 10, 11]])
 
             y1 = paddle.median(x)
-            # Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
-            #        [5.50000000])
+            # Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
+            #        5.50000000)
 
             y2 = paddle.median(x, axis=0)
             # Tensor(shape=[4], dtype=float32, place=Place(cpu), stop_gradient=True,
@@ -416,7 +415,7 @@ def median(x, axis=None, keepdim=False, name=None):
             -1,
             0,
             None,
-        ], 'when input 0D, axis can only be [-1, 0] or default None'
+        ], 'when input 0-D, axis can only be [-1, 0] or default None'
         is_flatten = True
 
     if axis is None:

python/paddle/tensor/to_string.py

@@ -181,8 +181,7 @@ def _format_tensor(var, summary, indent=0, max_width=0, signed=False):
     linewidth = DEFAULT_PRINT_OPTIONS.linewidth
 
     if len(var.shape) == 0:
-        # currently, shape = [], i.e., scaler tensor is not supported.
-        # If it is supported, it should be formatted like this.
+        # 0-D Tensor, whose shape = [], should be formatted like this.
         return _format_item(var, max_width, signed)
     elif len(var.shape) == 1:
         item_length = max_width + 2
@@ -291,7 +290,7 @@ def _format_dense_tensor(tensor, indent):
     if tensor.dtype == core.VarDesc.VarType.BF16:
         tensor = tensor.astype('float32')
 
-    # TODO(zhouwei): will remove 0D Tensor.numpy() hack
+    # TODO(zhouwei): will remove 0-D Tensor.numpy() hack
     np_tensor = tensor.numpy(False)
 
     if len(tensor.shape) == 0:

python/paddle/utils/cpp_extension/extension_utils.py

@@ -562,8 +562,6 @@ def normalize_extension_kwargs(kwargs, use_cuda=False):
                 extra_compile_args[compiler] = []
 
     if IS_WINDOWS:
-        # TODO(zhouwei): may append compile flags in future
-        pass
         # append link flags
         extra_link_args = kwargs.get('extra_link_args', [])
         extra_link_args.extend(MSVC_LINK_FLAGS)