array_ops error recti and others

mindspore/_checkparam.py

@@ -493,11 +493,12 @@ class Validator:
             if addition_error_info is None:
                 addition_error_info = ''
             type_str = (type(type_).__name__ if isinstance(type_, (tuple, list)) else "") + str(type_)
-            raise TypeError(f'For \'{prim_name}\', the type of `{arg_name}` should be subclass'
-                            f' of {", ".join((str(x) for x in template_types))}, but got {type_str}'
-                            f' {addition_error_info}. This message is only for reference. The supported data types'
-                            f' depend on the hardware that executes the operator'
-                            f' and it is a subset of the data types above.')
+            raise TypeError(f"For '{prim_name}', the type of '{arg_name}'"
+                            f" should be {'one of ' if len(template_types) > 1 else ''}"
+                            f"{', '.join((str(x) for x in template_types))}, but got {type_str}"
+                            f" {addition_error_info}. The supported data types depend on the hardware that"
+                            f" executes the operator, please refer the official api document to get"
+                            f" more information about the data type.")
 
     @staticmethod
     def check_valid_input(arg_name, arg_value, prim_name):

mindspore/nn/layer/activation.py

@@ -607,10 +607,11 @@ class PReLU(Cell):
                                  f"float32 when the 'w' is a tensor, but got {w.dtype}.")
             if len(w.shape) != 1 or w.shape[0] != channel:
                 raise ValueError(f"For '{self.cls_name}', the dimension of 'w' should be 1, and the elements number "
-                                 f"should be equal to the 'channel' when the 'w' is a tensor, but got 'w' shape {w}, "
-                                 f"the 'channel' {channel}.")
+                                 f"should be equal to the 'channel' when the 'w' is a tensor, "
+                                 f"but got 'w' shape {w.shape}, the 'channel' {channel}.")
         else:
-            raise TypeError(f"For '{self.cls_name}', the 'w' only supported float, list and tensor, but got {type(w)}.")
+            raise TypeError(f"For '{self.cls_name}', the 'w' only supported float, list and tensor, "
+                            f"but got {type(w).__name__}.")
         self.w = Parameter(w, name='a')
         self.prelu = P.PReLU()
         self.relu = P.ReLU()

mindspore/nn/layer/basic.py

@@ -313,7 +313,7 @@ class Dense(Cell):
         self.activation = get_activation(activation) if isinstance(activation, str) else activation
         if activation is not None and not isinstance(self.activation, (Cell, Primitive)):
             raise TypeError(f"For '{self.cls_name}', the 'activation' must be str or Cell or Primitive, but got "
-                            f"{type(activation)}.")
+                            f"{type(activation).__name__}.")
         self.activation_flag = self.activation is not None
 
     def construct(self, x):
@@ -800,7 +800,8 @@ class Pad(Cell):
         self.paddings = paddings
         Validator.check_string(self.mode, ["CONSTANT", "REFLECT", "SYMMETRIC"], 'mode', self.cls_name)
         if not isinstance(paddings, tuple):
-            raise TypeError(f"For '{self.cls_name}', the type of 'paddings' must be tuple, but got {type(paddings)}.")
+            raise TypeError(f"For '{self.cls_name}', the type of 'paddings' must be tuple, "
+                            f"but got {type(paddings).__name__}.")
         for item in paddings:
             if len(item) != 2:
                 raise ValueError(f"For '{self.cls_name}', the dimension of 'paddings' must be (n, 2), "
@@ -827,14 +828,15 @@ def bilinear(shape, size, scale, align_corners, prim_name=None):
     """Check input and calculate shape"""
     msg_prefix = f"For '{prim_name}', the" if prim_name else "The"
     if not isinstance(align_corners, bool):
-        raise TypeError(f"{msg_prefix} type of 'align_corners' should be boolean, but got {type(align_corners)}.")
+        raise TypeError(f"{msg_prefix} type of 'align_corners' should be boolean, "
+                        f"but got {type(align_corners).__name__}.")
     if size is None and scale is None:
         raise ValueError(f"{msg_prefix} 'size' and 'scale' both none.")
     if size is not None and scale is not None:
         raise ValueError(f"{msg_prefix} 'size' and 'scale' both not none.")
     if size is not None:
         if not isinstance(size, (tuple, list)):
-            raise ValueError(f"{msg_prefix} 'size' must be tuple or list, but got {type(size)}.")
+            raise ValueError(f"{msg_prefix} 'size' must be tuple or list, but got {type(size).__name__}.")
         Validator.check_int(len(size), 2, Rel.EQ, "size", "bilinear")
         Validator.check_int(size[0], 1, Rel.GE, "size[0]", "bilinear")
         Validator.check_int(size[1], 1, Rel.GE, "size[1]", "bilinear")

mindspore/nn/layer/embedding.py

@@ -299,8 +299,9 @@ class EmbeddingLookup(Cell):
             self.embeddinglookup.shard(((1, 1), indices_strategy))
         else:
             if is_auto_parallel:
-                Support_mode = ["field_slice", "table_row_slice", "table_column_slice", "batch_slice"]
-                validator.check_string(slice_mode, Support_mode, "slice_mode", self.cls_name)
+                support_mode = ["field_slice", "table_row_slice", "table_column_slice", "batch_slice"]
+                raise ValueError("For '{}', the 'slice_mode' must be in {}, "
+                                 "but got {}.".format(self.cls_name, support_mode, slice_mode))
         if self.cache_enable and not enable_ps:
             if parallel_mode != ParallelMode.STAND_ALONE:
                 raise ValueError(f"For '{self.cls_name}', parallel mode haven't supported cache enable yet.")

mindspore/nn/layer/lstm.py

@@ -46,7 +46,7 @@ def _check_input_dtype(input_dtype, param_name, allow_dtypes, cls_name):
 @constexpr
 def _check_input_3d(input_shape, param_name, func_name):
     if len(input_shape) != 3:
-        raise ValueError(f"For '{func_name}', the {param_name} should be 3d, but got the length of input_shape:"
+        raise ValueError(f"For '{func_name}', the '{param_name}' should be 3d, but got the length of input_shape:"
                          f" {len(input_shape)}.")
 
 

mindspore/nn/layer/normalization.py

@@ -180,7 +180,7 @@ class _BatchNorm(Cell):
             validator.check_int_range(rid, 0, rank_size, Rel.INC_LEFT, "rank id in process_groups")
             if rid in seen:
                 raise ValueError(f"For '{self.cls_name}', rank id in 'process_groups' should not be duplicated, "
-                                 f"but got {rid}.")
+                                 f"but got {process_groups}.")
             seen.add(rid)
 
     def _create_global_groups(self):
@@ -241,7 +241,7 @@ class _BatchNorm(Cell):
 def _channel_check(channel, num_channel, prim_name=None):
     msg_prefix = f"For '{prim_name}', the" if prim_name else "The"
     if channel != num_channel:
-        raise ValueError(f"{msg_prefix} channel should be equal with num_channel, but got channel: "
+        raise ValueError(f"{msg_prefix} channel should be equal to num_channel, but got channel: "
                          f"{channel}, num_channel: {num_channel}.")
 
 
@@ -888,7 +888,7 @@ class LayerNorm(Cell):
         super(LayerNorm, self).__init__()
         if not isinstance(normalized_shape, (tuple, list)):
             raise TypeError(f"For '{self.cls_name}', the type of 'normalized_shape' should be tuple[int] or list[int], "
-                            f"but got '{normalized_shape}' and the type is {type(normalized_shape)}.")
+                            f"but got {normalized_shape} and the type is {type(normalized_shape)}.")
         self.normalized_shape = normalized_shape
         self.begin_norm_axis = begin_norm_axis
         self.begin_params_axis = begin_params_axis
@@ -1037,7 +1037,7 @@ class InstanceNorm2d(Cell):
             val = args_dict[key]
             if not isinstance(val, (Tensor, numbers.Number, str, Initializer)):
                 raise TypeError(f"For '{self.cls_name}', the type of args_dict['{key}'] should be in "
-                                f"[Tensor, numbers.Number, str, Initializer], but got type {type(val)}.")
+                                f"[Tensor, numbers.Number, str, Initializer], but got type {type(val).__name__}.")
             if isinstance(val, Tensor) and val.dtype != mstype.float32:
                 raise TypeError(f"For '{self.cls_name}', the type of args_dict['{key}'] should be float32, "
                                 f"but got {val.dtype}.")

mindspore/nn/layer/quant.py

@@ -409,7 +409,7 @@ class FakeQuantWithMinMaxObserver(UniformQuantObserver):
         max_array = self._get_init_array(self.max_init)
         if not np.greater(max_array, min_array).all():
             raise ValueError(f"For '{self.cls_name}', the 'max_init' should be greater than 'min_init', "
-                             f"but got 'max_array': {max_array}, 'min_init': {min_init}.")
+                             f"but got 'max_init': {max_init}, 'min_init': {min_init}.")
         if self.mode == "DEFAULT":
             self._default_init(min_array, max_array)
         elif self.mode == "LEARNED_SCALE":
@@ -708,7 +708,7 @@ class Conv2dBnFoldQuantOneConv(Cell):
             self.padding = padding
         else:
             raise TypeError(f"For '{self.cls_name}', the type of 'padding' must be int/tuple(int), but got "
-                            f"{type(padding)}!")
+                            f"{type(padding).__name__}!")
         self.group = Validator.check_positive_int(group)
         self.eps = eps
         self.momentum = 1 - momentum
@@ -950,7 +950,7 @@ class Conv2dBnFoldQuant(Cell):
             self.padding = padding
         else:
             raise TypeError(f"For '{self.cls_name}', the type of 'padding' must be int/tuple(int), "
-                            f"but got {type(padding)}!")
+                            f"but got {type(padding).__name__}!")
         self.group = Validator.check_positive_int(group)
         self.eps = eps
         self.momentum = momentum
@@ -1162,7 +1162,7 @@ class Conv2dBnWithoutFoldQuant(Cell):
             self.padding = padding
         else:
             raise TypeError(f"For '{self.cls_name}', the type of 'padding' must be int/tuple(int), "
-                            f"but got {type(padding)}!")
+                            f"but got {type(padding).__name__}!")
         self.group = Validator.check_positive_int(group)
         self.bias_add = P.BiasAdd()
         if Validator.check_bool(has_bias):
@@ -1306,7 +1306,7 @@ class Conv2dQuant(Cell):
             self.padding = padding
         else:
             raise TypeError(f"For '{self.cls_name}', the type of 'padding' must be int/tuple(int), "
-                            f"but got {type(padding)}!")
+                            f"but got {type(padding).__name__}!")
         self.group = Validator.check_positive_int(group)
 
         weight_shape = [out_channels, in_channels // group, *self.kernel_size]

mindspore/nn/layer/thor_layer.py

@@ -217,7 +217,7 @@ class _ConvThor(Cell):
             self.padding = padding
         else:
             raise TypeError(f"For '{self.cls_name}', the type of 'padding' must be int/tuple(int), but got "
-                            f"{type(padding)}.")
+                            f"{type(padding).__name__}.")
 
         self.dilation = dilation
         self.group = Validator.check_positive_int(group)
@@ -766,7 +766,7 @@ class EmbeddingLookupThor(Cell):
                                  f"when 'slice_mode' is 'field_slice'.")
             if not isinstance(manual_shapes, tuple):
                 raise TypeError(f"For '{self.cls_name}', the type of 'manual_shapes' must be tuple(int), but got "
-                                f"type {type(manual_shapes)}.")
+                                f"type {type(manual_shapes).__name__}.")
             for dim in manual_shapes:
                 Validator.check_positive_int(dim, 'manual shape dim', self.cls_name)
             self.gatherv2.add_prim_attr("manual_split", manual_shapes)

mindspore/nn/layer/timedistributed.py

@@ -104,8 +104,8 @@ class TimeDistributed(Cell):
     def __init__(self, layer, time_axis, reshape_with_axis=None):
         """Initialize TimeDistributed."""
         if not isinstance(layer, (Cell, Primitive)):
-            raise TypeError(f"For '{self.cls_name}', the type of 'layer' should be mindspore.nn.Cell or "
-                            "mindspore.ops.Primitive instance, but got type: {type(layer)}.")
+            raise TypeError(f"For '{self.cls_name}', the 'layer' should be Cell or Primitive instance, "
+                            f"but got type: {type(layer).__name__}.")
         super(TimeDistributed, self).__init__()
         Validator.check_is_int(time_axis)
         if reshape_with_axis is not None:

mindspore/nn/loss/loss.py

@@ -665,8 +665,8 @@ def _check_weights(weight_shape, label_shape, prim_name=None):
     """Internal function, used to check whether the reduced shape meets the requirements."""
     msg_prefix = f'For \'{prim_name}\', the' if prim_name else "The"
     if weight_shape != label_shape:
-        raise ValueError(f"{msg_prefix} weight shape[0] should be equal to label.shape[1], "
-                         f"but got weight_shape: {weight_shape} and label_shape: {label_shape}.")
+        raise ValueError(f"{msg_prefix} weight_shape[0] should be equal to label_shape[1], "
+                         f"but got weight_shape[0]: {weight_shape} and label_shape[1]: {label_shape}.")
 
 
 class MultiClassDiceLoss(LossBase):

mindspore/ops/_utils/utils.py

@@ -64,7 +64,9 @@ def get_broadcast_shape(x_shape, y_shape, prim_name, shape_type=""):
             elif shape_type == "max_shape":
                 broadcast_shape_back.append(min(x_shape[i], y_shape[i]))
             else:
-                raise ValueError(f"For '{prim_name}', the x_shape {x_shape} and y_shape {y_shape} can not broadcast.")
+                raise ValueError(f"For '{prim_name}', x_shape and y_shape can broadcast means that "
+                                 f"x_shape[i] = 1 or -1 or y_shape[i] = 1 or -1 or x_shape[i] = y_shape[i], "
+                                 f"but got i: {i}, x_shape: {x_shape}, y_shape: {y_shape}.")
 
     broadcast_shape_front = y_shape[0: y_len - length] if length == x_len else x_shape[0: x_len - length]
     broadcast_shape = list(broadcast_shape_front) + broadcast_shape_back

mindspore/ops/composite/base.py

@@ -319,14 +319,13 @@ class GradOperation(GradOperation_):
     def __init__(self, get_all=False, get_by_list=False, sens_param=False):
         """Initialize GradOperation."""
         if not isinstance(get_all, bool):
-            raise TypeError(f"For 'GradOperation', the arg 'get_all' should be bool, but got {get_all} with type "
-                            f"{type(get_all)}")
+            raise TypeError(f"For 'GradOperation', the 'get_all' should be bool, but got {type(get_all).__name__}")
         if not isinstance(get_by_list, bool):
-            raise TypeError(f"For 'GradOperation', the arg 'get_by_list' should be bool, but got {get_by_list} with "
-                            f"type {type(get_by_list)}")
+            raise TypeError(f"For 'GradOperation', the 'get_by_list' should be bool, "
+                            f"but got {type(get_by_list).__name__}")
         if not isinstance(sens_param, bool):
-            raise TypeError(f"For 'GradOperation', the arg 'sens_param' should be bool, but got {sens_param} with type "
-                            f"{type(sens_param)}")
+            raise TypeError(f"For 'GradOperation', the 'sens_param' should be bool, "
+                            f"but got {type(sens_param).__name__}")
         self.get_all = get_all
         self.get_by_list = get_by_list
         self.sens_param = sens_param

mindspore/ops/composite/math_ops.py

@@ -488,7 +488,7 @@ def _check_axes_for_batch_dot(x1_shape, x2_shape, axes, prim_name=None):
             axes = [axes, axes]
     else:
         raise ValueError(f"{msg_prefix} type of axes must be one of those: int, tuple(int), list(int), "
-                         f"but got {type(axes)}.")
+                         f"but got {type(axes).__name__}.")
     return axes
 
 

mindspore/ops/operations/array_ops.py

@@ -17,12 +17,11 @@
 # ============================================================================
 
 """Operators for array."""
-
+from collections import Counter
 import copy
 import functools
 import itertools
 import numbers
-
 import numpy as np
 
 from mindspore import log as logger
@@ -87,7 +86,8 @@ class _ScatterOpDynamic(PrimitiveWithCheck):
     def _check_scatter_shape(self, x_shape, indices_shape, updates_shape, prim_name):
         # x_shape cannot be dynamic
         if np.any(np.array(x_shape) == -1):
-            raise ValueError(f"x does not support dynamic shape")
+            raise ValueError(f"For '{prim_name}', the 'input_x' does not support dynamic shape, "
+                             f"but got the shape of 'input_x' is {x_shape}.")
         # support indices and updates dynamic
         if np.any(np.array(indices_shape) == -1) or np.any(np.array(updates_shape) == -1):
             pass
@@ -492,7 +492,8 @@ class Reshape(PrimitiveWithInfer):
             validator.check_value_type("shape[%d]" % i, shp_i, [int], self.name)
             if shp_i == -1:
                 if neg_index != -1:
-                    raise ValueError(f'The shape can only has one -1 at most, but {shape_v}.')
+                    raise ValueError(f"For '{self.name}', the 'input_shape' can only has one -1 at most, "
+                                     f"but got {shape_v}.")
                 neg_index = i
             else:
                 dim_prod *= shp_i
@@ -514,25 +515,27 @@ class Reshape(PrimitiveWithInfer):
                 max_shape[neg_index] = int(max_arr_prod / dim_prod)
                 min_shape[neg_index] = int(min_arr_prod / dim_prod)
             else:
-                raise ValueError(f'For dynamic shape, Reshape must have neg index')
+                raise ValueError(f"For '{self.name}', the 'input_shape' must have -1 in the case of dynamic shape, "
+                                 f"but got {shape_v}.")
             out = {'shape': shape['value'],
                    'dtype': x['dtype'],
                    'value': None,
                    'max_shape': tuple(max_shape),
                    'min_shape': tuple(min_shape)}
         else:
-            if dim_prod <= 0 or arr_prod % dim_prod != 0:
-                raise ValueError(f'For \'{self.name}\' input_x\'s shape is {x_shp}, input_shape\'s value is {shape_v}.'
-                                 f'The product of input_x\'s shape should > 0, '
-                                 f'and can be divided by product of input_shape, but '
-                                 f'product of input_x\'s shape is {arr_prod}, product of input_shape is {dim_prod}.')
+            if dim_prod <= 0:
+                raise ValueError(f"For '{self.name}', the shape of 'input_x' is {x_shp}, "
+                                 f"the value of 'input_shape' is {shape_v}. "
+                                 f"The product of shape of 'input_shape' should > 0, but got {dim_prod}.")
             if neg_index != -1:
                 shape_v[neg_index] = int(arr_prod / dim_prod)
                 dim_prod *= shape_v[neg_index]
             if dim_prod != arr_prod:
-                raise ValueError(f'For \'{self.name}\' input_x\'s shape is {x_shp}, input_shape\'s value is {shape_v}.'
-                                 f'The product of input_x\'s shape should be equal to product of input_shape, but '
-                                 f'product of input_x\'s shape is {arr_prod}, product of input_shape is {dim_prod}.')
+                raise ValueError(f"For '{self.name}', the shape of 'input_x' is {x_shp}, "
+                                 f"the value of 'input_shape' value is {shape_v}. "
+                                 f"The product of the shape of 'input_x' should be equal to product of 'input_shape', "
+                                 f"but product of the shape of 'input_x' is {arr_prod} "
+                                 f", product of 'input_shape' is {dim_prod}.")
             value = None
             if x['value'] is not None:
                 value = Tensor(x['value'].asnumpy().reshape(shape_v))
@@ -675,7 +678,8 @@ class Squeeze(PrimitiveWithInfer):
             for a in axis:
                 validator.check_int_range(a, -ndim, ndim - 1, Rel.INC_BOTH, 'axis or its elements', self.name)
                 if x_shape[a] != 1:
-                    raise ValueError('Cannot select an axis to squeeze out which has size not equal to one.')
+                    raise ValueError(f"For '{self.name}', the shape of 'input_x' at {a} dimension should be 1, "
+                                     f"but got {x_shape[a]}.")
             ret = [x_shape[i] for i in range(ndim) if not (i in axis or (i - ndim) in axis)]
         return ret
 
@@ -1077,8 +1081,9 @@ class Split(PrimitiveWithCheck):
             # only validate when shape fully known
             output_valid_check = x_shape[self.axis] % self.output_num
             if output_valid_check != 0:
-                raise ValueError(f"x_shape[{self.axis}] {x_shape[self.axis]} must be divide exactly by"
-                                 f" output_num {self.output_num}")
+                raise ValueError(f"For '{self.name}', the shape of 'input_x' is {x_shape}, 'axis' is {self.axis}, "
+                                 f"the shape of 'input_x' in 'axis' {self.axis} is {x_shape[self.axis]}, "
+                                 f"which must be divide exactly by 'output_num': {self.output_num}.")
         size_splits = [x_shape[self.axis] // self.output_num] * self.output_num
         self.add_prim_attr('size_splits', size_splits)
 
@@ -1449,7 +1454,7 @@ class TupleToArray(PrimitiveWithInfer):
         for i, item in enumerate(x):
             validator.check_value_type(f"x[{i}]", item, [numbers.Number], self.name)
         if not all(isinstance(item, dtype) for item in x):
-            raise TypeError("For \'{self.name}\' all elements of input x must be have same type.")
+            raise TypeError(f"For \'{self.name}\', all elements of 'input_x' must be have same type.")
         if isinstance(x[0], int):
             ret = np.array(x, np.int32)
         else:
@@ -1591,14 +1596,14 @@ class InvertPermutation(PrimitiveWithInfer):
         x_shp = x['shape']
         x_value = x['value']
         if x_value is None:
-            raise ValueError(f'For \'{self.name}\' the input value must be const.')
+            raise ValueError(f"For '{self.name}', the value of 'input_x' can not be None, but got {x_value}.")
         validator.check_value_type("shape", x_shp, [tuple, list], self.name)
         if mstype.issubclass_(x['dtype'], mstype.tensor):
-            raise ValueError(f'For \'{self.name}\' the input value must be non-Tensor.')
+            raise ValueError(f"For \'{self.name}\', the value of 'input_x' must be non-Tensor, but got {x['dtype']}")
         for shp in x_shp:
             if shp:
                 x_rank = len(np.array(x_value, np.int64).shape)
-                raise ValueError(f'For \'{self.name}\' the rank of input must be 1, but got {x_rank}.')
+                raise ValueError(f"For \'{self.name}\', the length of 'input_x' must be 1, but got {x_rank}.")
         for i, value in enumerate(x_value):
             validator.check_value_type("input[%d]" % i, value, [int], self.name)
         z = [x_value[i] for i in range(len(x_value))]
@@ -1606,7 +1611,8 @@ class InvertPermutation(PrimitiveWithInfer):
 
         for i in range(1, len(z)):
             if z[i - 1] == z[i]:
-                raise ValueError(f"For {self.name}, {z[i]} is duplicated in the input.")
+                raise ValueError(f"For '{self.name}', the 'input_x' can not contain duplicate values, "
+                                 f"but got duplicated {z[i]} in the 'input_x'.")
         validator.check(f'value min', min(x_value), '', 0, Rel.EQ, self.name)
         validator.check(f'value max', max(x_value), '', len(x_value) - 1, Rel.EQ, self.name)
 
@@ -1949,7 +1955,8 @@ class Tile(PrimitiveWithInfer):
 
     def check_elim(self, base_tensor, multiplier):
         if (not isinstance(base_tensor, Tensor)) or (not isinstance(multiplier, tuple)):
-            raise TypeError("Expecting (Tensor, tuple), got: ({}, {})".format(base_tensor, multiplier))
+            raise TypeError(f"For '{self.name}', the type of ('input_x', 'multiples') should be (Tensor, tuple), "
+                            f"but got ({type(base_tensor).__name__}, {type(multiplier).__name__}).")
         if all(v == 1 for v in multiplier):
             return (True, base_tensor)
         return (False, None)
@@ -1970,8 +1977,8 @@ class Tile(PrimitiveWithInfer):
                 x_shp.insert(0, 1)
             multiples_w = multiples_v
         elif len_sub < 0:
-            raise ValueError(f'For \'{self.name}\' the length of multiples can not be smaller than '
-                             f'the length of dimension in input_x.')
+            raise ValueError(f"For '{self.name}', the length of 'multiples' can not be smaller than "
+                             f"the length of dimension in 'input_x'.")
         for i, a in enumerate(multiples_w):
             x_shp[i] *= a
         value = None
@@ -2073,7 +2080,9 @@ class UnsortedSegmentSum(PrimitiveWithInfer):
             output_min_shape = list(num_segments['min_value'])
         else:
             if isinstance(num_segments_type, type(mstype.tensor)):
-                raise ValueError("Num_segments only support int type when it is not a dynamic value")
+                raise ValueError(f"For '{self.name}', the dtype of 'num_segments' only support int type "
+                                 f"when it is not a dynamic value, but got type of 'num_segments': "
+                                 f"{num_segments_type}.")
             output_max_shape = [num_segments_v]
             output_min_shape = [num_segments_v]
         if 'max_shape' in x and 'min_shape' in x:
@@ -2727,8 +2736,8 @@ class Slice(PrimitiveWithInfer):
             validator.check_non_negative_int(begin_v[i], f'input begin[{i}]')
             if x_shape[i] < begin_v[i] + size_v[i]:
                 y = begin_v[i] + size_v[i]
-                raise ValueError("For '%s' slice shape can not bigger than origin shape %d, %d." %
-                                 (self.name, x_shape[i], y))
+                raise ValueError(f"For '{self.name}', the sliced shape can not greater than origin shape, but got "
+                                 f"sliced shape is {y}, and origin shape is {x_shape}.")
         return {'shape': size_v,
                 'dtype': x['dtype'],
                 'value': None}
@@ -2794,7 +2803,9 @@ class ReverseV2(PrimitiveWithInfer):
                 normalized_axis.append(v)
 
         if len(normalized_axis) != len(set(normalized_axis)):
-            raise ValueError('axis cannot contain duplicate dimensions.')
+            duplicated = [item for item, count in Counter(normalized_axis).items() if count > 1]
+            raise ValueError(f"For '{self.name}', the 'axis' cannot contain duplicate dimensions,"
+                             f" but got duplicated elements {duplicated}.")
 
         return x_shape
 
@@ -2913,8 +2924,9 @@ class Select(Primitive):
 def _compute_slicing_length(begin, end, stride, x_shape, i):
     """Computes the length of the slicing."""
     if i >= len(x_shape):
-        raise ValueError(f"For 'StridedSlice', When their is no new axis, the index length must be less or "
-                         f"equal than the dim of x.")
+        raise ValueError(f"For 'StridedSlice', the index length must be less than or equal to "
+                         f"the dimension of 'input_x' when there is no new axis, but got "
+                         f"the dimension of 'input_x': {len(x_shape)} and the index length: {i}.")
     x_dim = x_shape[i]
     if stride > 0:
         # When slicing forward, convert begin and end to positive numbers.
@@ -3143,15 +3155,16 @@ class StridedSlice(PrimitiveWithInfer):
         validator.check_value_type("strides", strides_v, [tuple], self.name)
 
         if tuple(filter(lambda x: not isinstance(x, int), begin_v + end_v + strides_v)):
-            raise TypeError(f"For {self.name}, both the begins, ends, and strides must be a tuple of int, "
-                            f"but got begins: {begin_v}, ends: {end_v}, strides: {strides_v}.")
+            raise TypeError(f"For {self.name}, both the 'begin', 'end', and 'strides' must be a tuple of int, "
+                            f"but got 'begin': {begin_v}, 'end': {end_v}, 'strides': {strides_v}.")
 
         if tuple(filter(lambda x: x == 0, strides_v)):
-            raise ValueError(f"For '{self.name}', the strides cannot contain 0, but got strides: {strides_v}.")
+            raise ValueError(f"For '{self.name}', the 'strides' cannot contain 0, but got 'strides': {strides_v}.")
 
         if len(end_v) != len(begin_v) or len(strides_v) != len(begin_v):
-            raise ValueError(f"For '{self.name}' the length of begin index: {begin_v}, end index: {end_v} and "
-                             f"strides: {strides_v} must be equal.")
+            raise ValueError(f"For '{self.name}', the length of 'begin' index and the length of 'end' index "
+                             f"must be the same as 'strides', but got the length of 'begin': {begin_v}, "
+                             f"'end' index: {end_v} and 'strides': {strides_v}.")
 
         ret_shape = self._compute_slicing_shape(x['shape'], begin_v, end_v, strides_v)
 
@@ -3219,9 +3232,9 @@ class StridedSlice(PrimitiveWithInfer):
                     continue
                 if j < len(shrink_axis_pos) and shrink_axis_pos[j] == '1':
                     if (not -x_shape[i] <= begin < x_shape[i]) or stride < 0:
-                        raise IndexError(f"For {self.name}, when shrink axis, the stride cannot be negative number, "
-                                         f"and begin should be in [-{x_shape[i]}, {x_shape[i]}), "
-                                         f"but got stride: {stride}, begin: {begin}.")
+                        raise IndexError(f"For '{self.name}', the 'strides' cannot be negative number and "
+                                         f"'begin' should be in [-{x_shape[i]}, {x_shape[i]}) when shrink axis, "
+                                         f"but got 'strides': {stride}, 'begin': {begin}.")
                     j += 1
                     i += 1
                     continue
@@ -3253,9 +3266,9 @@ class StridedSlice(PrimitiveWithInfer):
                     continue
                 if j < len(shrink_axis_pos) and shrink_axis_pos[j] == '1':
                     if (not -x_shape[i] <= begin < x_shape[i]) or stride < 0:
-                        raise ValueError(f"For {self.name}, when shrink axis, the stride cannot be negative number, "
-                                         f"and begin should be in [-{x_shape[i]}, {x_shape[i]}), "
-                                         f"but got stride: {stride}, begin: {begin}.")
+                        raise IndexError(f"For '{self.name}', the 'strides' cannot be negative number and "
+                                         f"'begin' should be in [-{x_shape[i]}, {x_shape[i]}) when shrink axis, "
+                                         f"but got 'strides': {stride}, 'begin': {begin}.")
                     j += 1
                     i += 1
                     continue
@@ -3369,8 +3382,8 @@ class DiagPart(PrimitiveWithInfer):
     def infer_shape(self, x_shape):
         if len(x_shape) % 2 != 0 or \
                 not x_shape:
-            raise ValueError(f"For \'{self.name}\' input rank must be non-zero and even, but got rank {len(x_shape)}, "
-                             f"with shapes {x_shape}")
+            raise ValueError(f"For \'{self.name}\', the rank of 'input_x' must be non-zero and even, "
+                             f"but got rank {len(x_shape)}, with shapes {x_shape}.")
         length = len(x_shape) // 2
         for i in range(length):
             validator.check('input_shape[i + len(input_shape)/2]', x_shape[i + length],
@@ -3533,7 +3546,9 @@ class ScatterNd(PrimitiveWithInfer):
 
         indices_shape, update_shape = indices["shape"], update["shape"]
         if indices_shape[0] != update_shape[0]:
-            raise ValueError(f'For \'{self.name}\' The indices_shape[0] and update_shape[0] must be equal.')
+            raise ValueError(f"For '{self.name}', the first shape of 'indices' must be the same as the first shape of "
+                             f"'updates', but got the first shape of 'indices': {indices_shape[0]}, "
+                             f"the first shape of 'updates': {update_shape[0]}.")
 
         return {'shape': shp,
                 'dtype': update['dtype'],
@@ -3704,15 +3719,22 @@ class TensorScatterUpdate(PrimitiveWithInfer):
 
     def infer_shape(self, input_x_shape, indices_shape, updates_shape):
         if len(indices_shape) < 2:
-            raise ValueError("For 'TensorScatterUpdate', rank of indices cannot be less than 2.")
+            raise ValueError(f"For '{self.name}', the dimension of 'indices' cannot be less than 2,"
+                             f" but got {len(indices_shape)}.")
 
         if indices_shape[-1] > len(input_x_shape):
-            raise ValueError("For 'TensorScatterUpdate', indices.shape[-1] cannot be greater than rank of input_x.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'indices' must be less than or equal to "
+                             f"the dimension of 'input_x', but got the "
+                             f"last dimension of 'indices': {indices_shape[-1]} and the dimension of 'input_x': "
+                             f"{len(indices_shape)}.")
 
         updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:]
         if updates_shape_check != updates_shape:
-            raise ValueError("For 'TensorScatterUpdate', input_x.shape, "\
-                             "indices.shape and updates.shape are incompatible.")
+            raise ValueError(f"For '{self.name}', the shape of 'update' must be equal to "
+                             f"the shape of updates_shape_check, but got the shape of 'update': {updates_shape},"
+                             f"and the shape of updates_shape_check: {updates_shape_check}. Please check the shape of "
+                             f"'indices' and 'input_x', they should be meeting followings formula:\n"
+                             f" updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:].")
 
         return input_x_shape
 
@@ -3782,14 +3804,22 @@ class TensorScatterAdd(PrimitiveWithInfer):
 
     def infer_shape(self, input_x_shape, indices_shape, updates_shape):
         if len(indices_shape) < 2:
-            raise ValueError("For 'TensorScatterAdd', rank of indices cannot be less than 2.")
+            raise ValueError(f"For '{self.name}', the dimension of 'indices' cannot be less than 2,"
+                             f" but got {len(indices_shape)}.")
 
         if indices_shape[-1] > len(input_x_shape):
-            raise ValueError("For 'TensorScatterAdd', indices.shape[-1] cannot be greater than rank of input_x.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'indices' must be less than or equal to "
+                             f"the dimension of 'input_x', but got the "
+                             f"last dimension of 'indices': {indices_shape[-1]} and the dimension of 'input_x': "
+                             f"{len(indices_shape)}.")
 
         updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:]
         if updates_shape_check != updates_shape:
-            raise ValueError("For 'TensorScatterAdd', input_x.shape, indices.shape and updates.shape are incompatible.")
+            raise ValueError(f"For '{self.name}', the shape of 'update' must be equal to "
+                             f"the shape of updates_shape_check, but got the shape of 'update': {updates_shape},"
+                             f"and the shape of updates_shape_check: {updates_shape_check}. Please check the shape of "
+                             f"'indices' and 'input_x', they should be meeting followings formula:\n"
+                             f" updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:].")
 
         return input_x_shape
 
@@ -4719,8 +4749,11 @@ class SpaceToDepth(PrimitiveWithInfer):
         out_shape = copy.deepcopy(x_shape)
         for i in range(2):
             if out_shape[i + 2] % self.block_size != 0:
-                raise ValueError(f'For \'{self.name}\' input shape[{i + 2}] {out_shape[i + 2]} should be '
-                                 f'fully divided by block_size {self.block_size}')
+                msg_prefix = "2nd" if i + 2 == 2 else "3rd"
+                raise ValueError(f"For '{self.name}', the shape of output with index {i + 2} must be divided "
+                                 f"exactly by 'block_size', but got the {msg_prefix} dimension "
+                                 f"of output: {out_shape[i + 2]} and "
+                                 f"'block_size': {self.block_size}.")
             out_shape[i + 2] //= self.block_size
 
         out_shape[1] *= self.block_size * self.block_size
@@ -4876,8 +4909,11 @@ class SpaceToBatch(PrimitiveWithInfer):
         for i in range(2):
             padded = out_shape[i + 2] + self.paddings[i][0] + self.paddings[i][1]
             if padded % self.block_size != 0:
-                raise ValueError(f'For \'{self.name}\' padded[{i}] {padded} should be divisible by '
-                                 f'block_size {self.block_size}')
+                msg_ndim = "2nd" if i + 2 == 2 else "3rd"
+                raise ValueError(f"For '{self.name}', the shape of the output tensor should be "
+                                 f"divisible by 'block_size', but got the {msg_ndim} dimension of output: {padded} and "
+                                 f"'block_size': {self.block_size}. Please check the official homepage "
+                                 f"for more information about the output tensor.")
             out_shape[i + 2] = padded // self.block_size
         out_shape[0] *= self.block_size * self.block_size
         return out_shape
@@ -4964,8 +5000,9 @@ class BatchToSpace(PrimitiveWithInfer):
             out_shape[i + 2] = x_block_prod - crops_sum
         block_size_prod = self.block_size * self.block_size
         if out_shape[0] % block_size_prod != 0:
-            raise ValueError(f'For \'{self.name}\' input_x dimension 0 {out_shape[0]}  should be divisible by '
-                             f'block_size_prod {block_size_prod}')
+            raise ValueError(f"For '{self.name}', the shape of output with index 0 must be divided exactly "
+                             f"by block_size_prod, but got the shape of output: {out_shape} and "
+                             f"block_size_prod: {block_size_prod}.")
         out_shape[0] = out_shape[0] // block_size_prod
         return out_shape
 
@@ -5066,8 +5103,11 @@ class SpaceToBatchND(PrimitiveWithInfer):
             padded = out_shape[i + offset] + self.paddings[i][0] + \
                      self.paddings[i][1]
             if padded % self.block_shape[i] != 0:
-                raise ValueError(f'For \'{self.name}\' padded[{i}] {padded} should be divisible by '
-                                 f'block_shape[{i}] {self.block_shape[i]}')
+                msg_ndim = "2nd" if i + 2 == 2 else "3rd"
+                raise ValueError(f"For '{self.name}', the 2nd and 3rd dimension of the output tensor should be "
+                                 f"divisible by 'block_shape', but got the {msg_ndim} dimension of output: {padded} "
+                                 f"and the {i} dimension block_shape: {self.block_shape}. Please check the "
+                                 f"official homepage for more information about the output tensor.")
             out_shape[i + offset] = padded // self.block_shape[i]
             block_shape_prod = block_shape_prod * self.block_shape[i]
         out_shape[0] *= block_shape_prod
@@ -5170,8 +5210,9 @@ class BatchToSpaceND(PrimitiveWithInfer):
             out_shape[i + offset] = x_block_prod - crops_sum
 
         if out_shape[0] % block_shape_prod != 0:
-            raise ValueError(f'For \'{self.name}\' input_x dimension 0 {out_shape[0]} should be divisible by '
-                             f'block_shape_prod {block_shape_prod}')
+            raise ValueError(f"For '{self.name}', the 0th dimension of the output tensor should be "
+                             f"divisible by block_shape_prod, but got 0th dimension of the output tensor: "
+                             f"{out_shape[0]} and the block_shape_prod: {block_shape_prod}.")
         out_shape[0] = out_shape[0] // block_shape_prod
         return out_shape
 
@@ -5317,8 +5358,7 @@ class Meshgrid(PrimitiveWithInfer):
     def __init__(self, indexing="xy"):
         """Initialize Meshgrid."""
         validator.check_value_type("indexing", indexing, (str), self.name)
-        if indexing not in ("xy", "ij"):
-            raise ValueError("indexing parameter must be either 'xy' or 'ij'")
+        validator.check_string(indexing.lower(), ["xy", "ij"], "indexing", self.name)
         self.indexing = indexing
 
     def infer_shape(self, x_shape):
@@ -5401,7 +5441,8 @@ class InplaceUpdate(PrimitiveWithInfer):
                         Rel.EQ, self.name)
         for i in self.indices:
             if i < 0 or i >= x_shape[0]:
-                raise ValueError(f'The value of indices must be in [0, {x_shape[0]}), but got {i}.')
+                raise ValueError(f"For '{self.name}', the value of indices must be in [0, {x_shape[0]}), "
+                                 f"but got {i}.")
         x_rank = len(x_shape)
         for idx in range(x_rank)[1:]:
             validator.check('v dim %d' % idx, v_shape[idx], "x dim %d" % idx, x_shape[idx], Rel.EQ, self.name)
@@ -5731,10 +5772,11 @@ class EmbeddingLookup(PrimitiveWithCheck):
         validator.check_subclass("offset", offset['dtype'], mstype.int_, self.name)
         indices_shp = indices['shape']
         if not indices_shp:
-            raise ValueError("'indices' should NOT be a scalar.")
+            raise ValueError(f"For '{self.name}', the 'input_indices' should not be a scalar, but got {indices_shp}.")
         params_shp = params['shape']
         if len(params_shp) > 2:
-            raise ValueError("The dimension of 'params' in EmbeddingLookup must <= 2, but got %d." % len(params_shp))
+            raise ValueError(f"For '{self.name}', the dimension of 'input_params' must <= 2, "
+                             f"but got {len(params_shp)}.")
 
 
 class GatherD(Primitive):
@@ -6014,9 +6056,11 @@ class SearchSorted(PrimitiveWithInfer):
 
     def infer_shape(self, sequence_shape, values_shape):
         if len(sequence_shape) != 1 and sequence_shape[:-1] != values_shape[:-1]:
-            raise ValueError(f"Sequence should be 1 dimensional or has all but the last dimension matching "
-                             f" the dimensions of values, but got sequence's dimensions: {sequence_shape} "
-                             f"and values' dimensions: {values_shape}.")
+            raise ValueError(f"For '{self.name}', the 'sequence' should be 1 dimensional or "
+                             f"all dimensions except the last dimension of 'sequence' "
+                             f"must be the same as all dimensions except the last dimension of 'values'. "
+                             f"but got dimension of 'sequence': {sequence_shape} "
+                             f"and dimension of 'values': {values_shape}.")
         return values_shape
 
     def infer_dtype(self, sequence_dtype, values_dtype):
@@ -6081,14 +6125,22 @@ class TensorScatterMax(PrimitiveWithInfer):
 
     def infer_shape(self, input_x_shape, indices_shape, updates_shape):
         if len(indices_shape) < 2:
-            raise ValueError("For 'TensorScatterMax', rank of indices cannot be less than 2.")
+            raise ValueError(f"For '{self.name}', the dimension of 'indices' cannot be less than 2,"
+                             f" but got {len(indices_shape)}.")
 
         if indices_shape[-1] > len(input_x_shape):
-            raise ValueError("For 'TensorScatterMax', indices.shape[-1] cannot be greater than rank of input_x.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'indices' must be less than or equal to "
+                             f"the dimension of 'input_x', but got the "
+                             f"last dimension of 'indices': {indices_shape[-1]} and the dimension of 'input_x': "
+                             f"{len(indices_shape)}.")
 
         updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:]
         if updates_shape_check != updates_shape:
-            raise ValueError("For 'TensorScatterMax', input_x.shape, indices.shape and updates.shape are incompatible.")
+            raise ValueError(f"For '{self.name}', the shape of 'update' must be equal to "
+                             f"the shape of updates_shape_check, but got the shape of 'update': {updates_shape},"
+                             f"and the shape of updates_shape_check: {updates_shape_check}. Please check the shape of "
+                             f"'indices' and 'input_x', they should be meeting followings formula:\n"
+                             f" updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:].")
 
         return input_x_shape
 
@@ -6157,14 +6209,22 @@ class TensorScatterMin(PrimitiveWithInfer):
 
     def infer_shape(self, input_x_shape, indices_shape, updates_shape):
         if len(indices_shape) < 2:
-            raise ValueError("For 'TensorScatterMin', rank of indices cannot be less than 2.")
+            raise ValueError(f"For '{self.name}', the dimension of 'indices' cannot be less than 2,"
+                             f" but got {len(indices_shape)}.")
 
         if indices_shape[-1] > len(input_x_shape):
-            raise ValueError("For 'TensorScatterMin', indices.shape[-1] cannot be greater than rank of input_x.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'indices' must be less than or equal to "
+                             f"the dimension of 'input_x', but got the "
+                             f"last dimension of 'indices': {indices_shape[-1]} and the dimension of 'input_x': "
+                             f"{len(indices_shape)}.")
 
         updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:]
         if updates_shape_check != updates_shape:
-            raise ValueError("For 'TensorScatterMin', input_x.shape, indices.shape and updates.shape are incompatible.")
+            raise ValueError(f"For '{self.name}', the shape of 'update' must be equal to "
+                             f"the shape of updates_shape_check, but got the shape of 'update': {updates_shape},"
+                             f"and the shape of updates_shape_check: {updates_shape_check}. Please check the shape of "
+                             f"'indices' and 'input_x', they should be meeting followings formula:\n"
+                             f" updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:].")
 
         return input_x_shape
 
@@ -6234,14 +6294,22 @@ class TensorScatterSub(PrimitiveWithInfer):
 
     def infer_shape(self, input_x_shape, indices_shape, updates_shape):
         if len(indices_shape) < 2:
-            raise ValueError("For 'TensorScatterSub', rank of indices cannot be less than 2.")
+            raise ValueError(f"For '{self.name}', the dimension of 'indices' cannot be less than 2,"
+                             f" but got {len(indices_shape)}.")
 
         if indices_shape[-1] > len(input_x_shape):
-            raise ValueError("For 'TensorScatterSub', indices.shape[-1] cannot be greater than rank of input_x.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'indices' must be less than or equal to "
+                             f"the dimension of 'input_x', but got the "
+                             f"last dimension of 'indices': {indices_shape[-1]} and the dimension of 'input_x': "
+                             f"{len(indices_shape)}.")
 
         updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:]
         if updates_shape_check != updates_shape:
-            raise ValueError("For 'TensorScatterSub', input_x.shape, indices.shape and updates.shape are incompatible.")
+            raise ValueError(f"For '{self.name}', the shape of 'update' must be equal to "
+                             f"the shape of updates_shape_check, but got the shape of 'update': {updates_shape},"
+                             f"and the shape of updates_shape_check: {updates_shape_check}. Please check the shape of "
+                             f"'indices' and 'input_x', they should be meeting followings formula:\n"
+                             f" updates_shape_check = indices_shape[:-1] + input_x_shape[indices_shape[-1]:].")
 
         return input_x_shape
 

mindspore/ops/operations/comm_ops.py

@@ -147,7 +147,7 @@ class AllReduce(PrimitiveWithInfer):
     def __init__(self, op=ReduceOp.SUM, group=GlobalComm.WORLD_COMM_GROUP):
         """Initialize AllReduce."""
         if not isinstance(op, type(ReduceOp.SUM)):
-            raise TypeError(f"For '{self.name}', the 'op' of AllReduce should be str, but got {type(op)}.")
+            raise TypeError(f"For '{self.name}', the 'op' of AllReduce should be str, but got {type(op).__name__}.")
         if not isinstance(_get_group(group), str):
             raise TypeError(f"For '{self.name}', the 'group' of AllReduce should be str, "
                             f"but got {type(_get_group(group))}.")
@@ -563,7 +563,7 @@ class Broadcast(PrimitiveWithInfer):
 
     def infer_dtype(self, x_dtype):
         if not isinstance(x_dtype, tuple):
-            raise TypeError(f"For '{self.name}', the 'input_x' should be a tuple, but got {type(x_dtype)}!")
+            raise TypeError(f"For '{self.name}', the 'input_x' should be a tuple, but got {type(x_dtype).__name__}!")
         for _ele in x_dtype:
             validator.check_tensor_dtype_valid('x', _ele, target_dtypes, self.name)
         return x_dtype

mindspore/ops/operations/debug_ops.py

@@ -227,7 +227,7 @@ class HistogramSummary(PrimitiveWithInfer):
         # In the summary, the histogram value should be a tensor whose shape is not [].
         if not v_shape:
             raise ValueError(f"For '{self.name}', the type of 'value' should be tensor, "
-                             f"and whose shape should not be [], but got {v_shape}.")
+                             f"its shape should not be [], but got {v_shape}.")
 
         return SUMMARY_RETURN_VALUE
 
@@ -343,7 +343,7 @@ class HookBackward(PrimitiveWithInfer):
         self.add_prim_attr("cell_id", cell_id)
         self.init_attrs["cell_id"] = cell_id
         if not isinstance(hook_fn, (FunctionType, MethodType)):
-            raise TypeError(f"For '{self.name}', the tye of 'hook_fn' should be python function, "
+            raise TypeError(f"For '{self.name}', the type of 'hook_fn' should be python function, "
                             f"but got {type(hook_fn)}.")
         self.register_hook(hook_fn)
         self.cell_id = cell_id

mindspore/ops/operations/math_ops.py

@@ -2366,7 +2366,7 @@ class Minimum(_MathBinaryOp):
         >>> # case 1 : same data type
         >>> x = Tensor(np.array([1.0, 5.0, 3.0]), mindspore.float32)
         >>> y = Tensor(np.array([4.0, 2.0, 6.0]), mindspore.float32)
-        >>> minimum = ops.Minimum
+        >>> minimum = ops.Minimum()
         >>> output = minimum(x, y)
         >>> print(output)
         [1. 2. 3.]
@@ -2540,7 +2540,7 @@ class Div(_MathBinaryOp):
         >>> print(output)
         [-2.  2.5  3.]
         >>> print(output.dtype)
-        Flaot32
+        Float32
     """
 
     def infer_value(self, x, y):

mindspore/ops/operations/nn_ops.py

@@ -1264,7 +1264,9 @@ class BatchNorm(PrimitiveWithInfer):
         validator.check_float_range(momentum, 0, 1, Rel.INC_BOTH, 'momentum', self.name)
         self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
         if context.get_context("device_target") != "GPU" and self.format == "NHWC":
-            raise ValueError("NHWC format only support in GPU target.")
+            raise ValueError(f"For '{self.name}', the \"NHWC\" format only support in GPU target, "
+                             f"but got the format is {self.format} and "
+                             f"the platform is {context.get_context('device_target')}.")
         self.add_prim_attr('data_format', self.format)
         self.init_prim_io_names(inputs=['x', 'scale', 'offset', 'mean', 'variance'],
                                 outputs=['y', 'batch_mean', 'batch_variance', 'reserve_space_1', 'reserve_space_2'])
@@ -1331,7 +1333,7 @@ class Conv2D(Primitive):
         mode (int): Modes for different convolutions. 0 Math convolutiuon, 1 cross-correlation convolution ,
                        2 deconvolution, 3 depthwise convolution. Default: 1.
         pad_mode (str): Specifies padding mode. The optional values are
-            "same", "valid", "pad". Default: "valid".
+            "same", "valid", "pad", not case sensitive. Default: "valid".
 
             - same: Adopts the way of completion. The height and width of the output will be the same as
               the input `x`. The total number of padding will be calculated in horizontal and vertical
@@ -1413,10 +1415,11 @@ class Conv2D(Primitive):
             validator.check_equal_int(len(pad), 4, 'pad size', self.name)
         self.add_prim_attr("pad", pad)
         self.padding = pad
-        self.pad_mode = validator.check_string(pad_mode, ['valid', 'same', 'pad'], 'pad_mode', self.name)
+        self.pad_mode = validator.check_string(pad_mode.lower(), ['valid', 'same', 'pad'], 'pad_mode', self.name)
 
         if pad_mode != 'pad' and pad != (0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', padding must be zero when pad_mode is '{pad_mode}'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be zero when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' and 'pad_mode' is {pad_mode}.")
         if self.pad_mode == 'pad':
             for item in pad:
                 validator.check_non_negative_int(item, 'pad item', self.name)
@@ -1424,7 +1427,9 @@ class Conv2D(Primitive):
         self.mode = validator.check_equal_int(mode, 1, 'mode', self.name)
         self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
         if context.get_context("device_target") != "GPU" and self.format == "NHWC":
-            raise ValueError("NHWC format only support in GPU target.")
+            raise ValueError(f"For '{self.name}', the \"NHWC\" format only support in GPU target, "
+                             f"but got the format is {self.format} "
+                             f"and platform is {context.get_context('device_target')}.")
         self.add_prim_attr('data_format', self.format)
         self.out_channel = validator.check_positive_int(out_channel, 'out_channel', self.name)
         self.group = validator.check_positive_int(group, 'group', self.name)
@@ -1454,7 +1459,7 @@ class DepthwiseConv2dNative(PrimitiveWithInfer):
         mode (int): Modes for different convolutions. 0 Math convolution, 1 cross-correlation convolution ,
                        2 deconvolution, 3 depthwise convolution. Default: 3.
         pad_mode (str): Specifies padding mode. The optional values are
-            "same", "valid", "pad". Default: "valid".
+            "same", "valid", "pad". Default: "valid", not case sensitive.
 
             - same: Adopts the way of completion. The height and width of the output will be the same as
               the input `x`. The total number of padding will be calculated in horizontal and vertical
@@ -1541,9 +1546,10 @@ class DepthwiseConv2dNative(PrimitiveWithInfer):
             validator.check_equal_int(len(pad), 4, 'pad size', self.name)
         self.add_prim_attr("pad", pad)
         self.padding = pad
-        self.pad_mode = validator.check_string(pad_mode, ['valid', 'same', 'pad'], 'pad_mode', self.name)
+        self.pad_mode = validator.check_string(pad_mode.lower(), ['valid', 'same', 'pad'], 'pad_mode', self.name)
         if pad_mode != 'pad' and pad != (0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', padding must be zero when pad_mode is '{pad_mode}'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be zero when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' is {pad} and 'pad_mode' is {pad_mode}")
         if self.pad_mode == 'pad':
             for item in pad:
                 validator.check_non_negative_int(item, 'pad item', self.name)
@@ -1563,7 +1569,7 @@ class DepthwiseConv2dNative(PrimitiveWithInfer):
         _, _, stride_h, stride_w = self.stride
         _, _, dilation_h, dilation_w = self.dilation
         if kernel_size_n != 1:
-            raise ValueError(f"The batch of input weight should be 1, but got {kernel_size_n}")
+            raise ValueError(f"For '{self.name}', the batch of input weight should be 1, but got {kernel_size_n}")
         if self.pad_mode == "valid":
             h_out = math.ceil((x_shape[2] - dilation_h * (kernel_size_h - 1)) / stride_h)
             w_out = math.ceil((x_shape[3] - dilation_w * (kernel_size_w - 1)) / stride_w)
@@ -1630,7 +1636,9 @@ class _Pool(PrimitiveWithInfer):
         self.is_maxpoolwithargmax = (self.name == "MaxPoolWithArgmax")
         self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
         if context.get_context("device_target") != "GPU" and self.format == "NHWC":
-            raise ValueError("NHWC format only support in GPU target.")
+            raise ValueError(f"For '{self.name}', the \"NHWC\" format only support in GPU target, "
+                             f"but got the format is {self.format} and "
+                             f"the platform is {context.get_context('device_target')}.")
         if not self.is_maxpoolwithargmax:
             self.add_prim_attr('data_format', self.format)
 
@@ -1666,8 +1674,8 @@ class _Pool(PrimitiveWithInfer):
 
         for shape_value in out_shape:
             if shape_value <= 0:
-                raise ValueError(f"For '{self.name}', the kernel size is not valid, "
-                                 f"the input shape: {x_shape}, strides shape: {self.strides}.")
+                raise ValueError(f"For '{self.name}', the each element of the output shape must be larger than 0, "
+                                 f"but got one of shape in output shape is {shape_value}.")
         return out_shape
 
     def infer_dtype(self, x_dtype):
@@ -1925,10 +1933,11 @@ class MaxPool3D(PrimitiveWithInfer):
         if len(self.pad_list) == 3:
             self.pad_list = (pad_list[0], pad_list[0], pad_list[1], pad_list[1], pad_list[2], pad_list[2])
         if len(self.pad_list) != 3 and len(self.pad_list) != 6:
-            raise ValueError(f"For `maxpool3d` attr 'pad_list' should be an positive int number or a tuple of "
-                             f"three or six positive int numbers, but got `{len(self.pad_list)}` numbers.")
+            raise ValueError(f"For '{self.name}', attr 'pad_list' should be an positive int number or a tuple of "
+                             f"three or six positive int numbers, but got {len(self.pad_list)} numbers.")
         if self.pad_mode != 'CALCULATED' and self.pad_list != (0, 0, 0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', when pad_list is not 0, pad_mode should be set as 'pad'.")
+            raise ValueError(f"For '{self.name}', the 'pad_list' must be zero when 'pad_mode' is not \"CALCULATED\", "
+                             f"but got 'pad_list' is {self.pad_list} and 'pad_mode' is {pad_mode}.")
         if self.pad_mode == 'CALCULATED':
             for item in self.pad_list:
                 validator.check_non_negative_int(item, 'pad_list item', self.name)
@@ -2069,7 +2078,8 @@ class Conv2DBackpropInput(Primitive):
             and width of the 2D convolution window. Single int means the value is for both the height and the width of
             the kernel. A tuple of 2 ints means the first value is for the height and the other is for the
             width of the kernel.
-        pad_mode (str): Modes to fill padding. It could be "valid", "same", or "pad". Default: "valid".
+        pad_mode (str): Modes to fill padding. It could be "valid", "same", or "pad", not case sensitive.
+            Default: "valid".
         pad (Union[int, tuple[int]]): The pad value to be filled. Default: 0. If `pad` is an integer, the paddings of
                     top, bottom, left and right are the same, equal to pad. If `pad` is a tuple of four integers, the
                     padding of top, bottom, left and right equal to pad[0], pad[1], pad[2], and pad[3] correspondingly.
@@ -2146,7 +2156,9 @@ class Conv2DBackpropInput(Primitive):
         self.add_prim_attr('kernel_size', self.kernel_size)
         self.format = validator.check_string(data_format, ['NCHW', 'NHWC'], 'format', self.name)
         if context.get_context("device_target") != "GPU" and self.format == "NHWC":
-            raise ValueError("NHWC format only support in GPU target.")
+            raise ValueError(f"For '{self.name}', the \"NHWC\" format only support in GPU target, "
+                             f"but got the format is {self.format} and "
+                             f"the platform is {context.get_context('device_target')}.")
         self.add_prim_attr('data_format', self.format)
         self.stride = _check_positive_int_or_tuple('stride', stride, self.name, allow_four=True, ret_four=True)
         self.stride = _update_attr_by_format(self.stride, self.format)
@@ -2161,9 +2173,10 @@ class Conv2DBackpropInput(Primitive):
             validator.check_equal_int(len(pad), 4, 'pad size', self.name)
         self.add_prim_attr("pad", pad)
         self.padding = pad
-        self.pad_mode = validator.check_string(pad_mode, ['valid', 'same', 'pad'], 'pad_mode', self.name)
+        self.pad_mode = validator.check_string(pad_mode.lower(), ['valid', 'same', 'pad'], 'pad_mode', self.name)
         if pad_mode != 'pad' and pad != (0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', padding must be zero when pad_mode is '{pad_mode}'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be zero when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' is {pad} and 'pad_mode' is {pad_mode}.")
         if self.pad_mode == 'pad':
             for item in pad:
                 validator.check_non_negative_int(item, 'pad item', self.name)
@@ -2284,7 +2297,9 @@ class BiasAdd(Primitive):
         self.init_prim_io_names(inputs=['x', 'b'], outputs=['output'])
         self.format = validator.check_string(data_format, ['NCHW', 'NHWC', 'NCDHW'], 'format', self.name)
         if context.get_context("device_target") != "GPU" and self.format == "NHWC":
-            raise ValueError("NHWC format only support in GPU target.")
+            raise ValueError(f"For '{self.name}', the \"NHWC\" format only support in GPU target, "
+                             f"but got the format is {self.format} and "
+                             f"the platform is {context.get_context('device_target')}.")
         self.add_prim_attr('data_format', self.format)
 
 
@@ -2388,7 +2403,8 @@ class NLLLoss(PrimitiveWithInfer):
         \end{array}\right.
 
     Args:
-        reduction (str): Apply specific reduction method to the output: 'none', 'mean', 'sum'. Default: "mean".
+        reduction (str): Apply specific reduction method to the output: 'none', 'mean', 'sum', not case sensitive.
+            Default: "mean".
 
     Inputs:
         - **logits** (Tensor) - Input logits, with shape :math:`(N, C)`. Data type only support float32 or float16.
@@ -2432,7 +2448,7 @@ class NLLLoss(PrimitiveWithInfer):
     def __init__(self, reduction="mean"):
         """Initialize NLLLoss"""
         self.init_prim_io_names(inputs=['x', 'target', "weight"], outputs=['loss'])
-        self.reduction = validator.check_string(reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+        self.reduction = validator.check_string(reduction.lower(), ['none', 'sum', 'mean'], 'reduction', self.name)
         self.add_prim_attr('reduction', self.reduction)
 
     def infer_shape(self, x_shape, t_shape, w_shape):
@@ -2640,8 +2656,8 @@ class ApplyMomentum(PrimitiveWithInfer):
     @prim_attr_register
     def __init__(self, use_nesterov=False, use_locking=False, gradient_scale=1.0):
         """Initialize ApplyMomentum."""
-        self.use_nesterov = validator.check_bool(use_nesterov)
-        self.use_locking = validator.check_bool(use_locking)
+        self.use_nesterov = validator.check_bool(use_nesterov, "use_nesterov", self.name)
+        self.use_locking = validator.check_bool(use_locking, "use_locking", self.name)
         validator.check_value_type('gradient_scale', gradient_scale, [float], self.name)
         self.init_prim_io_names(inputs=['variable', 'accumulation', 'learning_rate', 'gradient', 'momentum'],
                                 outputs=['output'])
@@ -2746,7 +2762,8 @@ class SoftMarginLoss(Primitive):
         \text{loss}(x, y) = \sum_i \frac{\log(1 + \exp(-y[i]*x[i]))}{\text{x.nelement}()}
 
     Args:
-        reduction (str): Apply specific reduction method to the output: 'none', 'mean', 'sum'. Default: "mean".
+        reduction (str): Apply specific reduction method to the output: 'none', 'mean', 'sum', not case sensitive.
+            Default: "mean".
 
     Inputs:
         - **logits** (Tensor) - Predict data. Data type must be float16 or float32.
@@ -2778,7 +2795,7 @@ class SoftMarginLoss(Primitive):
     def __init__(self, reduction="mean"):
         """Initialize SoftMarginLoss"""
         self.init_prim_io_names(inputs=['predict', 'label'], outputs=['loss'])
-        self.reduction = validator.check_string(reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+        self.reduction = validator.check_string(reduction.lower(), ['none', 'sum', 'mean'], 'reduction', self.name)
 
 
 class L2Loss(PrimitiveWithInfer):
@@ -3000,7 +3017,8 @@ class SGD(PrimitiveWithCheck):
         """Initialize SGD."""
         validator.check_value_type("nesterov", nesterov, [bool], self.name)
         if nesterov and dampening != 0:
-            raise ValueError(f"Nesterov need zero dampening!")
+            raise ValueError(f"For '{self.name}', the 'dampening' must be 0 when 'nesterov' is True, "
+                             f"but got 'dampening' is {dampening} and 'nesterov' is {nesterov}.")
         self.init_prim_io_names(inputs=['parameters', 'gradient', 'learning_rate', 'accum', 'momentum', 'stat'],
                                 outputs=['output'])
         self.add_prim_attr('side_effect_mem', True)
@@ -3128,7 +3146,8 @@ class ApplyRMSProp(PrimitiveWithInfer):
 
     def infer_value(self, var, mean_square, moment, learning_rate, grad, decay, momentum, epsilon):
         if decay is None or momentum is None or epsilon is None:
-            raise ValueError(f"For {self.name}, decay, momentum, epsilon must be const.")
+            raise ValueError(f"For '{self.name}', 'decay', 'momentum' and 'epsilon' can not be None, "
+                             f"but got 'decay': {decay}, 'momentum': {momentum} and 'epsilon':{epsilon}.")
 
 
 class ApplyCenteredRMSProp(PrimitiveWithInfer):
@@ -3362,7 +3381,8 @@ class L2Normalize(PrimitiveWithInfer):
         self.add_prim_attr('axis', axis)
         self.init_attrs['axis'] = axis
         if len(axis) != 1:
-            raise TypeError("The length of axis must be 1, later will support multiple axis!")
+            raise TypeError(f"For '{self.name}', the dimension of 'axis' must be 1, but got {len(axis)}, "
+                            f"later will support multiple axis!")
         self.axis = axis
 
     def infer_shape(self, input_x):
@@ -3842,17 +3862,18 @@ class PReLU(PrimitiveWithInfer):
         input_x_dim = len(input_x_shape)
         if input_x_dim in (0, 1):
             if context.get_context("device_target") == "Ascend":
-                raise ValueError(f"For '{self.name}', the 0-D or 1-D 'input_x' is not supported on Ascend.")
+                raise ValueError(f"For '{self.name}', the dimension of 'x' can not be 0-D or 1-D when the platform is "
+                                 f"\"Ascend\", but got dimension of 'x' is {input_x_dim}.")
             channel_num = 1
         else:
             channel_num = input_x_shape[1]
 
         weight_dim = len(weight_shape)
         if weight_dim != 1:
-            raise ValueError(f"For '{self.name}', the weight dimension should be 1, while got {weight_dim}.")
+            raise ValueError(f"For '{self.name}', the dimension of 'x' should be 1, while got {weight_dim}.")
         if weight_shape[0] != 1 and weight_shape[0] != channel_num:
-            raise ValueError(f"For '{self.name}', the weight shape should be (1,) or "
-                             f"matched with input channel ({channel_num},), but got {weight_shape}")
+            raise ValueError(f"For '{self.name}', the first dimension of 'weight' should be (1,) or "
+                             f"it should be equal to number of channels: {channel_num}, but got {weight_shape}")
         return input_x_shape
 
     def infer_dtype(self, input_x_dtype, weight_dtype):
@@ -4055,8 +4076,8 @@ class BCEWithLogitsLoss(PrimitiveWithInfer):
     and the third method is to calculate the sum of all losses.
 
     Args:
-        reduction (str): Type of reduction to be applied to loss. The optional values are 'mean', 'sum', and 'none'.
-            If 'none', do not perform reduction. Default:'mean'.
+        reduction (str): Type of reduction to be applied to loss. The optional values are 'mean', 'sum', and 'none',
+             not case sensitive. If 'none', do not perform reduction. Default:'mean'.
 
     Inputs:
         - **logits** (Tensor) - Input logits. Data type must be float16 or float32.
@@ -4095,36 +4116,36 @@ class BCEWithLogitsLoss(PrimitiveWithInfer):
     @prim_attr_register
     def __init__(self, reduction='mean'):
         """Initialize BCEWithLogitsLoss"""
-        self.reduction = validator.check_string(reduction, ['none', 'sum', 'mean'], 'reduction', self.name)
+        self.reduction = validator.check_string(reduction.lower(), ['none', 'sum', 'mean'], 'reduction', self.name)
 
-    def infer_shape(self, predict, target, weight, pos_weight):
-        validator.check('predict_shape', predict, 'target_shape', target, Rel.EQ, self.name)
+    def infer_shape(self, logits, label, weight, pos_weight):
+        validator.check('logits_shape', logits, 'label_shape', label, Rel.EQ, self.name)
         reversed_weight_shape = tuple(reversed(weight))
-        reversed_target = tuple(reversed(predict))
+        reversed_label = tuple(reversed(logits))
         for i, v in enumerate(reversed_weight_shape):
-            if v not in (reversed_target[i], 1):
-                raise ValueError(f"For {self.name}, shapes can not broadcast. "
-                                 f"predict: {tuple(predict)}, weight shape {tuple(weight)}.")
+            if v not in (reversed_label[i], 1):
+                raise ValueError(f"For {self.name}, the shapes of 'logits' and 'weight' can not broadcast. "
+                                 f"'logits': {tuple(logits)}, 'weight' shape {tuple(weight)}.")
 
         reversed_pos_shape = tuple(reversed(pos_weight))
-        reversed_target = tuple(reversed(predict))
+        reversed_label = tuple(reversed(logits))
         for i, v in enumerate(reversed_pos_shape):
-            if v not in (reversed_target[i], 1):
-                raise ValueError(f"For {self.name}, shapes can not broadcast. "
-                                 f"predict: {tuple(predict)}, pos_weight shape {tuple(pos_weight)}.")
+            if v not in (reversed_label[i], 1):
+                raise ValueError(f"For {self.name}, the shapes of 'logits' and 'pos_weight' can not broadcast. "
+                                 f"'logits': {tuple(logits)}, 'pos_weight' shape {tuple(pos_weight)}.")
 
         if self.reduction in ('mean', 'sum'):
             shape = []
         else:
-            shape = predict
+            shape = logits
         return shape
 
-    def infer_dtype(self, predict, target, weight, pos_weight):
-        validator.check_tensor_dtype_valid('predict dtype', predict, [mstype.float16, mstype.float32], self.name)
-        validator.check_tensor_dtype_valid('target dtype', target, [mstype.float16, mstype.float32], self.name)
+    def infer_dtype(self, logits, label, weight, pos_weight):
+        validator.check_tensor_dtype_valid('logits dtype', logits, [mstype.float16, mstype.float32], self.name)
+        validator.check_tensor_dtype_valid('label dtype', label, [mstype.float16, mstype.float32], self.name)
         validator.check_tensor_dtype_valid('weight dtype', weight, [mstype.float16, mstype.float32], self.name)
         validator.check_tensor_dtype_valid('pos_weight dtype', pos_weight, [mstype.float16, mstype.float32], self.name)
-        return predict
+        return logits
 
 
 class Pad(PrimitiveWithInfer):
@@ -4170,17 +4191,19 @@ class Pad(PrimitiveWithInfer):
         """Initialize Pad"""
         self.init_prim_io_names(inputs=['x'], outputs=['y'])
         if not isinstance(paddings, tuple):
-            raise TypeError('Paddings must be tuple type.')
+            raise TypeError(f"For '{self.name}', the type of 'paddings' must be tuple, "
+                            f"but got {type(paddings)}.")
         for item in paddings:
             if len(item) != 2:
-                raise ValueError('The shape of paddings must be (n, 2).')
+                raise ValueError(f"For '{self.name}', the shape of paddings must be (n, 2), "
+                                 f"but got {item}.")
         self.paddings = paddings
 
     def infer_shape(self, x_shape):
         validator.check_int(len(self.paddings), len(x_shape), Rel.EQ, 'paddings.shape', self.name)
         paddings = np.array(self.paddings)
         if not np.all(paddings >= 0):
-            raise ValueError('All elements of paddings must be >= 0.')
+            raise ValueError(f"For '{self.name}', all elements of paddings must be >= 0.")
         y_shape = ()
         for i in range(int(paddings.size / 2)):
             y_shape += ((x_shape[i] + paddings[i, 0] + paddings[i, 1]),)
@@ -4272,13 +4295,14 @@ class MirrorPad(PrimitiveWithInfer):
         paddings_size = paddings_value.size
         validator.check_int(paddings_size, len(x_shape) * 2, Rel.EQ, 'paddings.shape', self.name)
         if not np.all(paddings_value >= 0):
-            raise ValueError('All elements of paddings must be >= 0.')
+            raise ValueError(f"For '{self.name}', all elements of 'paddings' must be >= 0.")
         adjust = 0
         if self.mode == 'SYMMETRIC':
             adjust = 1
         for i in range(0, int(paddings_size / 2)):
             if (paddings_value[i, 0] >= x_shape[i] + adjust) or (paddings_value[i, 1] >= x_shape[i] + adjust):
-                raise ValueError('At least one dim has too high a padding value for this input and mode')
+                raise ValueError(f"For '{self.name}', both paddings[D, 0] and paddings[D, 1] must be no greater than "
+                                 f"the dimension corresponding to 'x'.")
         y_shape = ()
         for i in range(0, int(paddings_size / 2)):
             y_shape += ((x_shape[i] + paddings_value[i, 0] + paddings_value[i, 1]),)
@@ -5307,7 +5331,7 @@ class KLDivLoss(PrimitiveWithInfer):
 
     Args:
         reduction (str): Specifies the reduction to be applied to the output.
-            Its value must be one of 'none', 'mean', 'sum'. Default: 'mean'.
+            Its value must be one of 'none', 'mean', 'sum', not case sensitive. Default: 'mean'.
 
     Inputs:
         - **logits** (Tensor) - The input Tensor. The data type must be float32.
@@ -5345,7 +5369,7 @@ class KLDivLoss(PrimitiveWithInfer):
     @prim_attr_register
     def __init__(self, reduction='mean'):
         """Initialize KLDivLoss."""
-        self.reduction = validator.check_string(reduction, ['none', 'mean', 'sum'], 'reduction', self.name)
+        self.reduction = validator.check_string(reduction.lower(), ['none', 'mean', 'sum'], 'reduction', self.name)
 
     def infer_shape(self, x_shape, y_shape):
         validator.check('x_shape', x_shape, 'y_shape', y_shape, Rel.EQ, self.name)
@@ -5389,7 +5413,7 @@ class BinaryCrossEntropy(PrimitiveWithInfer):
 
     Args:
         reduction (str): Specifies the reduction to be applied to the output.
-            Its value must be one of 'none', 'mean', 'sum'. Default: 'mean'.
+            Its value must be one of 'none', 'mean', 'sum', not case sensitive. Default: 'mean'.
 
     Inputs:
         - **logits** (Tensor) - The input Tensor. The data type must be float16 or float32,
@@ -5432,7 +5456,7 @@ class BinaryCrossEntropy(PrimitiveWithInfer):
     @prim_attr_register
     def __init__(self, reduction='mean'):
         """Initialize BinaryCrossEntropy."""
-        self.reduction = validator.check_string(reduction, ['none', 'mean', 'sum'], 'reduction', self.name)
+        self.reduction = validator.check_string(reduction.lower(), ['none', 'mean', 'sum'], 'reduction', self.name)
 
     def infer_shape(self, x_shape, y_shape, weight_shape):
         validator.check('x_shape', x_shape, 'y_shape', y_shape, Rel.EQ, self.name)
@@ -6002,7 +6026,6 @@ class SparseApplyAdagrad(PrimitiveWithInfer):
     @prim_attr_register
     def __init__(self, lr, update_slots=True, use_locking=False):
         """Initialize SparseApplyAdagrad."""
-        validator.check_value_type("lr", lr, [float], self.name)
         validator.check_is_float(lr, "lr", self.name)
         validator.check_value_type("update_slots", update_slots, [bool], self.name)
         validator.check_value_type("use_locking", use_locking, [bool], self.name)
@@ -7674,14 +7697,15 @@ class DynamicRNN(PrimitiveWithInfer):
         validator.check_int(len(h_shape), 3, Rel.EQ, "h_shape", self.name)
         validator.check_int(len(c_shape), 3, Rel.EQ, "c_shape", self.name)
         if seq_shape is not None:
-            raise ValueError(f"For {self.name}, seq_shape should be None.")
+            raise ValueError(f"For '{self.name}', the dimension of 'seq_length' should be None, but got {seq_shape}.")
 
         num_step, batch_size, input_size = x_shape
         hidden_size = w_shape[-1] // 4
 
         validator.check("b_shape[-1]", b_shape[-1], "w_shape[-1]", w_shape[-1], Rel.EQ, self.name)
         if w_shape[-1] % 4 != 0:
-            raise ValueError(f"For {self.name}, w_shape[-1] should multiple of 4.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'w' should be a multiple of 4, "
+                             f"but got {w_shape[-1]}.")
         validator.check("w_shape[0]", w_shape[0], "input_size + hidden_size",
                         input_size + hidden_size, Rel.EQ, self.name)
         validator.check("b_shape[0]", b_shape[0], "w_shape[1]", w_shape[1], Rel.EQ, self.name)
@@ -7841,7 +7865,8 @@ class DynamicGRUV2(PrimitiveWithInfer):
         num_step, batch_size, input_size = x_shape
         hidden_size = winput_shape[-1] // 3
         if winput_shape[-1] % 3 != 0:
-            raise ValueError(f"For {self.name}, weight_input_shape[-1] should multiple of 3.")
+            raise ValueError(f"For '{self.name}', the last dimension of 'w' should be a multiple of 3, "
+                             f"but got {winput_shape[-1]}.")
 
         self.placeholder_index = [3, 4, 5]
         if binput_shape is not None:
@@ -7854,7 +7879,8 @@ class DynamicGRUV2(PrimitiveWithInfer):
                             "3 * hidden_shape", [3 * hidden_size], Rel.EQ, self.name)
             self.placeholder_index.remove(4)
         if seq_shape is not None:
-            raise ValueError(f"For {self.name}, seq_shape should be None.")
+            raise ValueError(f"For '{self.name}', the dimension of 'seq_length' should be None, "
+                             f"but got {seq_shape}.")
 
         validator.check_int(len(h_shape), 2, Rel.EQ, "init_h shape rank", self.name)
         validator.check("init_h_shape[0]", h_shape[0], "batch_size", batch_size, Rel.EQ, self.name)
@@ -8036,8 +8062,8 @@ class AvgPool3D(Primitive):
         strides (Union[int, tuple[int]]): The distance of kernel moving, an int number that represents
             the depth, height and width of movement are both strides, or a tuple of three int numbers that
             represent depth, height and width of movement respectively. Default: 1.
-        pad_mode (str): The optional value for pad mode, is "same", "valid", "pad", not case sensitive.
-            Default: "valid".
+        pad_mode (str): The optional value for pad mode, is "SAME", "VALID", "PAD", not case sensitive.
+            Default: "VALID".
 
             - same: Adopts the way of completion. The depth, height and width of the output will be the same as
               the input. The total number of padding will be calculated in depth, horizontal and vertical
@@ -8106,7 +8132,7 @@ class AvgPool3D(Primitive):
         if isinstance(pad, int):
             pad = (pad,) * 6
         if len(pad) != 6:
-            raise ValueError(f"For `AvgPool3D` attr 'pad' should be an positive int number or a tuple of "
+            raise ValueError(f"For '{self.name}', attr 'pad' should be an positive int number or a tuple of "
                              f"six positive int numbers, but got `{len(pad)}`.")
         self.pad_list = pad
         self.add_prim_attr('pad_list', self.pad_list)
@@ -8115,7 +8141,8 @@ class AvgPool3D(Primitive):
         self.add_prim_attr('pad_mode', self.pad_mode)
 
         if self.pad_mode != 'PAD' and pad != (0, 0, 0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', when pad is not 0, pad_mode should be set as 'pad'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be (0, 0, 0, 0, 0, 0) when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' is {pad} and 'pad_mode' is {self.pad_mode}.")
         if self.pad_mode == 'PAD':
             for item in pad:
                 validator.check_non_negative_int(item, 'pad or item of pad', self.name)
@@ -8250,7 +8277,7 @@ class Conv3D(PrimitiveWithInfer):
         if isinstance(pad, int):
             pad = (pad,) * 6
         if len(pad) != 6:
-            raise ValueError(f"For `conv3d` attr 'pad' should be an positive int number or a tuple of "
+            raise ValueError(f"For '{self.name}', attr 'pad' should be an positive int number or a tuple of "
                              f"six positive int numbers, but got `{len(pad)}`.")
         self.add_prim_attr("pad", pad)
         self.padding = pad
@@ -8259,7 +8286,8 @@ class Conv3D(PrimitiveWithInfer):
         self.add_prim_attr('pad_mode', self.pad_mode)
 
         if self.pad_mode != 'pad' and pad != (0, 0, 0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', when pad is not 0, pad_mode should be set as 'pad'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be (0, 0, 0, 0, 0, 0) when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' is {pad} and 'pad_mode' is {self.pad_mode}.")
         if self.pad_mode == 'pad':
             for item in pad:
                 validator.check_non_negative_int(item, 'pad item', self.name)
@@ -8277,7 +8305,7 @@ class Conv3D(PrimitiveWithInfer):
         validator.check_equal_int(len(w_shape), 5, "weight rank", self.name)
         validator.check_equal_int(len(x_shape), 5, "x rank", self.name)
         if b_shape is not None:
-            raise ValueError("Bias currently only support None.")
+            raise ValueError(f"For '{self.name}', the 'bias' currently only support None, but got {b_shape}.")
         validator.check(f"x_shape[1] // group", x_shape[1] // self.group, "w_shape[1]", w_shape[1], Rel.EQ, self.name)
         validator.check('out_channel', self.out_channel, 'w_shape[0]', w_shape[0], Rel.EQ, self.name)
         validator.check('kernel_size', self.kernel_size, 'w_shape[1:4]', tuple(w_shape[2:]), Rel.EQ, self.name)
@@ -8443,7 +8471,9 @@ class Conv3DBackpropInput(PrimitiveWithInfer):
 
         self.pad_mode = validator.check_string(pad_mode.lower(), ['valid', 'same', 'pad'], 'pad_mode', self.name)
         if self.pad_mode != 'pad' and self.pad_list != (0, 0, 0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', when pad is not 0, pad_mode should be set as 'pad'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be (0, 0, 0, 0, 0, 0) "
+                             f"when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' is {self.pad_list} and 'pad_mode' is {self.pad_mode}.")
         if self.pad_mode == 'pad':
             for item in pad:
                 validator.check_non_negative_int(item, 'pad item', self.name)
@@ -8526,8 +8556,8 @@ class CTCLossV2(Primitive):
 
     Args:
         blank (int): The blank label. Default: 0.
-        reduction (string): Apply specific reduction method to the output. Currently only support 'none'.
-          Default: "none".
+        reduction (string): Apply specific reduction method to the output. Currently only support 'none',
+            not case sensitive. Default: "none".
         zero_infinity (bool): Whether to set infinite loss and correlation gradient to zero. Default: False.
 
     Inputs:
@@ -8557,7 +8587,7 @@ class CTCLossV2(Primitive):
         self.add_prim_attr("blank", blank)
         validator.check_value_type("reduction", reduction, [str], self.name)
         self.reduction = self.reduction.lower()
-        validator.check_string(self.reduction, ['none'], 'reduction', self.name)
+        validator.check_string(self.reduction.lower(), ['none'], 'reduction', self.name)
         self.add_prim_attr("reduction", self.reduction)
         validator.check_value_type("zero_infinity", zero_infinity, [bool], self.name)
         self.add_prim_attr("zero_infinity", zero_infinity)
@@ -8640,7 +8670,7 @@ class Conv3DTranspose(PrimitiveWithInfer):
             other is for the width of the kernel.
         mode (int): Modes for different convolutions. Default is 1. It is currently not used.
         pad_mode (str): Specifies padding mode. The optional values are
-            "same", "valid", "pad". Default: "valid".
+            "same", "valid", "pad", not case sensitive. Default: "valid".
 
             - same: Adopts the way of completion. The depth, height and width of the output will be the same as
               the input. The total number of padding will be calculated in depth, horizontal and vertical
@@ -8736,7 +8766,7 @@ class Conv3DTranspose(PrimitiveWithInfer):
         if isinstance(pad, int):
             pad = (pad,) * 6
         if len(pad) != 6:
-            raise ValueError(f"For `Conv3DTranspose` attr 'pad' should be an positive int number or a tuple of "
+            raise ValueError(f"For '{self.name}', attr 'pad' should be an positive int number or a tuple of "
                              f"six positive int numbers, but got `{len(pad)}`.")
         self.pad_list = pad
         validator.check_value_type('pad_mode', pad_mode, [str], self.name)
@@ -8744,7 +8774,8 @@ class Conv3DTranspose(PrimitiveWithInfer):
         self.add_prim_attr('pad_mode', self.pad_mode)
 
         if self.pad_mode != 'pad' and pad != (0, 0, 0, 0, 0, 0):
-            raise ValueError(f"For '{self.name}', when pad is not 0, pad_mode should be set as 'pad'.")
+            raise ValueError(f"For '{self.name}', the 'pad' must be (0, 0, 0, 0, 0, 0) when 'pad_mode' is not \"pad\", "
+                             f"but got 'pad' is {pad} and 'pad_mode' is {self.pad_mode}.")
 
         if self.pad_mode == 'pad':
             for item in self.pad_list:
@@ -8760,7 +8791,9 @@ class Conv3DTranspose(PrimitiveWithInfer):
                                                      allow_five=False, ret_five=True, greater_zero=False)
         output_padding = (self.output_padding[2], self.output_padding[3], self.output_padding[4])
         if self.pad_mode != 'pad' and output_padding != (0, 0, 0):
-            raise ValueError(f"For '{self.name}', when output_padding is not 0, pad_mode should be set as 'pad'.")
+            raise ValueError(f"For '{self.name}', the 'output_padding' must be (0, 0, 0) "
+                             f"when 'pad_mode' is not \"pad\", "
+                             f"but got 'output_padding' is {output_padding} and 'pad_mode' is {self.pad_mode}.")
         validator.check_int_range(self.kernel_size[0] * self.kernel_size[1] * self.kernel_size[2], 1, 343, Rel.INC_BOTH,
                                   'The product of height, width and depth of kernel_size belonging [1, 343]', self.name)
         validator.check_int_range(self.stride[0] * self.stride[1] * self.stride[2], 1, 343, Rel.INC_BOTH,
@@ -8777,7 +8810,7 @@ class Conv3DTranspose(PrimitiveWithInfer):
     def __infer__(self, x, w, b=None):
         args = {'x': x['dtype'], 'w': w['dtype']}
         if b is not None:
-            raise ValueError("Bias currently only support None.")
+            raise ValueError(f"For '{self.name}', the 'bias' currently only support None, but got {b}.")
         valid_dtypes = [mstype.float16, mstype.float32]
         validator.check_tensors_dtypes_same_and_valid(args, valid_dtypes, self.name)
 

mindspore/ops/operations/other_ops.py

@@ -552,10 +552,11 @@ class CheckBprop(PrimitiveWithInfer):
         return xshapes
 
     def infer_dtype(self, xdtypes, ydtypes):
-        validator.check_value_type('grads', xdtypes, (tuple,), self.name)
-        validator.check_value_type('params', ydtypes, (tuple,), self.name)
+        tips = f'Bprop of {self.prim_to_check}'
+        validator.check_value_type('grads', xdtypes, (tuple,), tips)
+        validator.check_value_type('params', ydtypes, (tuple,), tips)
         if len(xdtypes) < len(ydtypes):
-            raise ValueError(f"For '{self.name}', the size of 'xdtypes' should not be less than {len(ydtypes)},"
+            raise ValueError(f"{tips}, the size of output should be {len(ydtypes)},"
                              f" but got {len(xdtypes)}.")
         checking_range = len(ydtypes)
         for i in range(checking_range):
@@ -565,11 +566,11 @@ class CheckBprop(PrimitiveWithInfer):
                 continue
             if isinstance(ydtype, mstype.function_type):
                 if not isinstance(xdtype, mstype.env_type_type):
-                    raise TypeError(f"For '{self.name}', the dtype of {i}th 'xdtypes' should be {mstype.env_type_type},"
+                    raise TypeError(f"{tips}, the dtype of {i}th output should be {mstype.env_type_type},"
                                     f" but got {xdtype}.")
                 continue
             if xdtype != ydtype:
-                raise TypeError(f"For '{self.name}', the shape of {i}th 'xdtypes' should be {ydtype},"
+                raise TypeError(f"{tips}, the dtype of {i}th output should be {ydtype},"
                                 f" but got {xdtype}.")
         return xdtypes
 

mindspore/ops/operations/random_ops.py

@@ -69,7 +69,7 @@ class StandardNormal(PrimitiveWithInfer):
     def __infer__(self, shape):
         shape_v = shape["value"]
         if shape_v is None:
-            raise ValueError(f"For '{self.name}', the 'shape' cannot be None, but got {shape}.")
+            raise ValueError(f"For '{self.name}', the 'shape' cannot be None.")
         Validator.check_value_type("shape", shape_v, [tuple], self.name)
         for i, shape_i in enumerate(shape_v):
             Validator.check_positive_int(shape_i, f'shape[{i}]', self.name)
@@ -126,7 +126,7 @@ class StandardLaplace(PrimitiveWithInfer):
     def __infer__(self, shape):
         shape_v = shape["value"]
         if shape_v is None:
-            raise ValueError(f"For '{self.name}', the 'shape' cannot be None, but got {shape}.")
+            raise ValueError(f"For '{self.name}', the 'shape' cannot be None.")
         Validator.check_value_type("shape", shape_v, [tuple], self.name)
         for i, shape_i in enumerate(shape_v):
             Validator.check_positive_int(shape_i, f'shape[{i}]', self.name)
@@ -189,7 +189,7 @@ class Gamma(PrimitiveWithInfer):
     def __infer__(self, shape, alpha, beta):
         shape_v = shape["value"]
         if shape_v is None:
-            raise ValueError(f"For '{self.name}', the 'shape' cannot be None, but got {shape}.")
+            raise ValueError(f"For '{self.name}', the 'shape' cannot be None.")
         Validator.check_value_type("shape", shape_v, [tuple], self.name)
         for i, shape_i in enumerate(shape_v):
             Validator.check_positive_int(shape_i, f'shape[{i}]', self.name)
@@ -253,7 +253,7 @@ class Poisson(PrimitiveWithInfer):
     def __infer__(self, shape, mean):
         shape_v = shape["value"]
         if shape_v is None:
-            raise ValueError(f"For '{self.name}', the 'shape' cannot be None, but got {shape}.")
+            raise ValueError(f"For '{self.name}', the 'shape' cannot be None.")
         Validator.check_value_type("shape", shape_v, [tuple], self.name)
         for i, shape_i in enumerate(shape_v):
             Validator.check_positive_int(shape_i, f'shape[{i}]', self.name)
@@ -325,7 +325,7 @@ class UniformInt(PrimitiveWithInfer):
     def __infer__(self, shape, minval, maxval):
         shape_v = shape["value"]
         if shape_v is None:
-            raise ValueError(f"For '{self.name}', the 'shape' cannot be None, but got {shape}.")
+            raise ValueError(f"For '{self.name}', the 'shape' cannot be None.")
         Validator.check_value_type("shape", shape_v, [tuple], self.name)
         for i, shape_i in enumerate(shape_v):
             Validator.check_positive_int(shape_i, f'shape[{i}]', self.name)
@@ -564,7 +564,7 @@ class Multinomial(PrimitiveWithInfer):
         Validator.check_tensor_dtype_valid('inputs', inputs['dtype'], [mstype.float32], self.name)
         num_samples_value = num_samples["value"]
         if num_samples_value is None:
-            raise ValueError(f"For '{self.name}', the 'num_samples' cannot be None, but got {num_samples}.")
+            raise ValueError(f"For '{self.name}', the 'num_samples' cannot be None.")
         Validator.check_value_type("num_samples", num_samples_value, (int,), self.name)
         Validator.check_positive_int(num_samples_value, "num_samples")
         y_shape = (num_samples_value,)

tests/st/ops/gpu/test_reverse_v2_op.py

@@ -107,9 +107,9 @@ def test_reverse_v2_invalid_axis():
     with pytest.raises(ValueError) as info:
         reverse_v2_net = ReverseV2Net((0, 1, 2, 1))
         _ = reverse_v2_net(x)
-    assert "axis cannot contain duplicate dimensions" in str(info.value)
+    assert "'axis' cannot contain duplicate dimensions" in str(info.value)
 
     with pytest.raises(ValueError) as info:
         reverse_v2_net = ReverseV2Net((-2, -1, 3))
         _ = reverse_v2_net(x)
-    assert "axis cannot contain duplicate dimensions" in str(info.value)
+    assert "'axis' cannot contain duplicate dimensions" in str(info.value)

tests/st/pynative/test_tensor_getitem.py

@@ -1204,7 +1204,7 @@ def test_tensor_slice_reduce_out_of_bounds_neg():
     net = NetWork()
     with pytest.raises(IndexError) as ex:
         net(input_tensor)
-    assert "begin should be in [-6, 6), but got stride: 1, begin: -7." in str(ex.value)
+    assert "'begin' should be in [-6, 6) when shrink axis, but got 'strides': 1, 'begin': -7." in str(ex.value)
 
 
 @pytest.mark.level1
@@ -1226,7 +1226,7 @@ def test_tensor_slice_reduce_out_of_bounds_positive():
     net = NetWork()
     with pytest.raises(IndexError) as ex:
         net(input_tensor)
-    assert "begin should be in [-6, 6), but got stride: 1, begin: 6." in str(ex.value)
+    assert "'begin' should be in [-6, 6) when shrink axis, but got 'strides': 1, 'begin': 6." in str(ex.value)
 
 
 @pytest.mark.level0

tests/ut/python/optimizer/test_auto_grad.py

@@ -386,7 +386,7 @@ def test_grad_args_type_error1():
     try:
         GradNetWrtX(Net())(x, y)
     except TypeError as e:
-        assert "For 'GradOperation', the arg 'get_all' should be bool, but got" in str(e)
+        assert "For 'GradOperation', the 'get_all' should be bool, but got" in str(e)
 
 
 def test_grad_args_type_error2():
@@ -412,7 +412,7 @@ def test_grad_args_type_error2():
     try:
         GradNetWrtX(Net())(x, y)
     except TypeError as e:
-        assert "For 'GradOperation', the arg 'get_by_list' should be bool, but got" in str(e)
+        assert "For 'GradOperation', the 'get_by_list' should be bool, but got" in str(e)
 
 
 def test_grad_args_type_error3():
@@ -438,7 +438,7 @@ def test_grad_args_type_error3():
     try:
         GradNetWrtX(Net())(x, y)
     except TypeError as e:
-        assert "For 'GradOperation', the arg 'sens_param' should be bool, but got" in str(e)
+        assert "For 'GradOperation', the 'sens_param' should be bool, but got" in str(e)
 
 
 def test_grad_net_is_none():

tests/ut/python/pipeline/parse/test_invert.py

@@ -53,11 +53,11 @@ def test_invert_int_tensor():
     context.set_context(mode=context.PYNATIVE_MODE)
     with pytest.raises(TypeError) as err:
         net(input_x)
-    assert "For 'LogicalNot or '~' operator', the type of `x` should be subclass of Tensor[Bool], " \
+    assert "For 'LogicalNot or '~' operator', the type of 'x' should be Tensor[Bool], " \
            "but got Tensor[Int32]" in str(err.value)
 
     context.set_context(mode=context.GRAPH_MODE)
     with pytest.raises(TypeError) as err:
         net(input_x)
-    assert "For 'LogicalNot or '~' operator', the type of `x` should be subclass of Tensor[Bool], " \
+    assert "For 'LogicalNot or '~' operator', the type of 'x' should be Tensor[Bool], " \
            "but got Tensor[Int32]" in str(err.value)