mindspore 2.4.1__cp311-cp311-manylinux1_x86_64.whl → 2.5.0__cp311-cp311-manylinux1_x86_64.whl

mindspore/nn/layer/pooling.py

@@ -18,23 +18,21 @@ from __future__ import absolute_import
 from mindspore.ops import operations as P
 from mindspore.ops import functional as F
 import mindspore.ops as ops
-from mindspore.ops.function.nn_func import avg_pool2d_ext
 from mindspore._checkparam import _check_3d_int_or_tuple
 from mindspore import _checkparam as validator
 from mindspore.ops.primitive import constexpr, _primexpr
 from mindspore.common.tensor import Tensor
 import mindspore.context as context
 from mindspore.common import dtype as mstype
-from mindspore.ops.operations.nn_ops import AdaptiveMaxPool2D
-from mindspore.ops.operations.nn_ops import AdaptiveMaxPool3D, AdaptiveAvgPool3D
-from mindspore.ops.auto_generate.gen_ops_prim import MaxPoolWithIndices, MaxPoolWithMask
 from mindspore.nn.cell import Cell
 from mindspore._c_expression import MSContext
+from mindspore.ops.auto_generate import avg_pool1d_ext
+
 
 __all__ = ['AvgPool3d', 'MaxPool3d', 'AvgPool2d', 'MaxPool2d', 'AvgPool1d', 'MaxPool1d', 'FractionalMaxPool2d',
            'FractionalMaxPool3d', 'AdaptiveAvgPool1d', 'AdaptiveMaxPool1d', 'AdaptiveMaxPool2d', 'AdaptiveMaxPool3d',
            'AdaptiveAvgPool2d', 'AdaptiveAvgPool3d', 'MaxUnpool1d', 'MaxUnpool2d', 'MaxUnpool3d', 'LPPool1d',
-           'LPPool2d', 'AvgPool2dExt', 'MaxPool2dExt']
+           'LPPool2d', 'AvgPool2dExt', 'MaxPool2dExt', 'AvgPool1dExt']
 
 
 class _PoolNd(Cell):
@@ -297,6 +295,9 @@ class MaxPool3d(_PoolNd):
         \max_{l=0, \ldots, d_{ker}-1} \max_{m=0, \ldots, h_{ker}-1} \max_{n=0, \ldots, w_{ker}-1}
         \text{input}(N_i, C_j, s_0 \times d + l, s_1 \times h + m, s_2 \times w + n)
 
+    .. note::
+        For Atlas training series products, this interface is not supported.
+
     Args:
         kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value,
             is an int number or a single element tuple that represents depth, height and width of the kernel, or a tuple
@@ -686,9 +687,11 @@ class MaxPool2dExt(Cell):
         self.return_indices = return_indices
         strides = stride if (stride is not None) else kernel_size
         if return_indices:
-            self.max_pool_func_ = MaxPoolWithIndices(kernel_size, strides, padding, dilation, ceil_mode)
+            self.max_pool_func_ = ops.auto_generate.gen_ops_prim.MaxPoolWithIndices(kernel_size, strides, padding,
+                                                                                    dilation, ceil_mode)
         else:
-            self.max_pool_func_ = MaxPoolWithMask(kernel_size, strides, padding, dilation, ceil_mode)
+            self.max_pool_func_ = ops.auto_generate.gen_ops_prim.MaxPoolWithMask(kernel_size, strides, padding,
+                                                                                 dilation, ceil_mode)
 
     def construct(self, input):
         out, indices = self.max_pool_func_(input)
@@ -1018,6 +1021,40 @@ class AvgPool3d(_PoolNd):
         return out
 
 
+class AvgPool1dExt(Cell):
+    r"""
+    Applies a 1D average pooling over an input Tensor which can be regarded as
+    a composition of 2D input planes.
+
+    For details, please refer to :func:`mindspore.mint.nn.functional.avg_pool1d`.
+
+    Supported Platforms:
+        ``Ascend``
+
+    Examples:
+        >>> import numpy as np
+        >>> from mindspore import Tensor, nn
+        >>> from mindspore import dtype as mstype
+        >>> input = Tensor(np.arange(1 * 3 * 4).reshape(1, 3, 4), mstype.float32)
+        >>> net = nn.AvgPool1dExt(kernel_size=2, stride=1)
+        >>> output = net(input)
+        >>> print(output.shape)
+        (1, 3, 3)
+    """
+    def __init__(self, kernel_size, stride=None, padding=0, ceil_mode=False,
+                 count_include_pad=True):
+        super().__init__()
+        self.kernel_size = kernel_size
+        self.stride = stride
+        self.padding = padding
+        self.ceil_mode = ceil_mode
+        self.count_include_pad = count_include_pad
+
+    def construct(self, input):
+        return avg_pool1d_ext(input, self.kernel_size, self.stride, self.padding,
+                              self.ceil_mode, self.count_include_pad)
+
+
 class AvgPool2dExt(Cell):
     r"""
     Applies a 2D average pooling over an input Tensor which can be regarded as
@@ -1032,16 +1069,11 @@ class AvgPool2dExt(Cell):
         >>> import numpy as np
         >>> from mindspore import Tensor, nn
         >>> from mindspore import dtype as mstype
-        >>> x = Tensor(np.arange(1 * 3 * 3 * 4).reshape(1, 3, 3, 4), mstype.float32)
-        >>> m =  nn.AvgPool2dExt(x, kernel_size=2, stride=1)
-        >>> output = m(x)
-        >>> print(output)
-        [[[[ 2.5   3.5   4.5]
-           [ 6.5   7.5   8.5]]
-          [[14.5  15.5  16.5]
-           [18.5  19.5  20.5]]
-          [[26.5  27.5  28.5]
-           [30.5  31.5  32.5]]]]
+        >>> input = Tensor(np.arange(1 * 3 * 3 * 4).reshape(1, 3, 3, 4), mstype.float32)
+        >>> net = nn.AvgPool2dExt(kernel_size=2, stride=1)
+        >>> output = net(input)
+        >>> print(output.shape)
+        (1, 3, 2, 3)
     """
     def __init__(self, kernel_size, stride=None, padding=0, ceil_mode=False,
                  count_include_pad=True, divisor_override=None):
@@ -1054,8 +1086,8 @@ class AvgPool2dExt(Cell):
         self.divisor_override = divisor_override
 
     def construct(self, input):
-        return avg_pool2d_ext(input, self.kernel_size, self.stride, self.padding,
-                              self.ceil_mode, self.count_include_pad, self.divisor_override)
+        return ops.function.nn_func.avg_pool2d_ext(input, self.kernel_size, self.stride, self.padding,
+                                                   self.ceil_mode, self.count_include_pad, self.divisor_override)
 
 
 class AvgPool2d(_PoolNd):
@@ -1129,7 +1161,7 @@ class AvgPool2d(_PoolNd):
         TypeError: If `kernel_size` or `strides` is neither int nor tuple.
         ValueError: If `pad_mode` is not ``"valid"`` , ``"same"`` or ``"pad"`` with not case sensitive.
         ValueError: If `data_format` is neither ``'NCHW'`` nor ``'NHWC'``.
-        ValueError: If `padding`, `ceil_mode`, `count_include_pad`, or `divisor_override` is used
+        ValueError: If `padding`, `ceil_mode`, `count_include_pad`, or `divisor_override` is used,
             or `pad_mode` is ``"pad"`` when `data_format` is 'NHWC'.
         ValueError: If `kernel_size` or `strides` is less than 1.
         ValueError: If length of `padding` tuple/list is not 1 or 2.
@@ -1594,7 +1626,7 @@ class AdaptiveAvgPool3d(Cell):
     def __init__(self, output_size):
         """Initialize AdaptiveAvgPool3d."""
         super(AdaptiveAvgPool3d, self).__init__()
-        self.adaptive_avg_pool3d = AdaptiveAvgPool3D(output_size)
+        self.adaptive_avg_pool3d = ops.AdaptiveAvgPool3D(output_size)
 
     def construct(self, input):
         return self.adaptive_avg_pool3d(input)
@@ -1766,7 +1798,7 @@ class AdaptiveMaxPool2d(Cell):
         """Initialize AdaptiveMaxPool2d."""
         super(AdaptiveMaxPool2d, self).__init__()
         validator.check_value_type('return_indices', return_indices, [bool], self.cls_name)
-        self.adaptive_max_pool2d = AdaptiveMaxPool2D(output_size)
+        self.adaptive_max_pool2d = ops.AdaptiveMaxPool2D(output_size)
         self.return_indices = return_indices
 
     def construct(self, input):
@@ -1825,7 +1857,7 @@ class AdaptiveMaxPool3d(Cell):
             output_size = (output_size, output_size, output_size)
         self.output_size = Tensor(output_size, dtype=mstype.int32)
         self.return_indices = return_indices
-        self.adaptive_max_pool3d = AdaptiveMaxPool3D()
+        self.adaptive_max_pool3d = ops.AdaptiveMaxPool3D()
 
     def construct(self, input):
         output = self.adaptive_max_pool3d(input, self.output_size)

mindspore/nn/layer/rnn_cells.py

@@ -340,7 +340,7 @@ class GRUCell(RNNCellBase):
     :math:`r` is reset gate. :math:`z` is update gate. :math:`n` is n-th layer. For instance,
     :math:`W_{ir}, b_{ir}` are the weight and bias used to transform from input :math:`x` to :math:`r`.
     Details can be found in paper
-    `Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
+    `Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
     <https://aclanthology.org/D14-1179.pdf>`_.
 
     Args:

mindspore/nn/layer/rnns.py

@@ -237,6 +237,7 @@ class _DynamicGRUCPUGPU(Cell):
                 h_0.view(1, *h_0.shape),
                 weights.astype(x.dtype)
             )
+
             if seq_length is not None:
                 h_n = get_hidden(output, seq_length)
                 mask = sequence_mask(seq_length, x.shape[0])
@@ -687,7 +688,7 @@ class GRU(_RNNBase):
     are learnable weights between the output and the input in the formula. For instance,
     :math:`W_{ir}, b_{ir}` are the weight and bias used to transform from input :math:`x` to :math:`r`.
     Details can be found in paper
-    `Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
+    `Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
     <https://aclanthology.org/D14-1179.pdf>`_.
 
     Note:

mindspore/nn/layer/timedistributed.py

@@ -15,8 +15,8 @@
 """Time Distributed."""
 from __future__ import absolute_import
 
+from mindspore import ops
 from mindspore.ops.primitive import constexpr, Primitive, _primexpr
-from mindspore.ops import Reshape, Transpose, Stack, Unstack
 from mindspore.common import Tensor
 from mindspore import _checkparam as Validator
 from mindspore.nn.cell import Cell
@@ -116,8 +116,8 @@ class TimeDistributed(Cell):
         self.layer = layer
         self.time_axis = time_axis
         self.reshape_with_axis = reshape_with_axis
-        self.transpose = Transpose()
-        self.reshape = Reshape()
+        self.transpose = ops.Transpose()
+        self.reshape = ops.Reshape()
 
     def construct(self, inputs):
         _check_data(isinstance(inputs, Tensor), self.cls_name)
@@ -143,7 +143,7 @@ class TimeDistributed(Cell):
             outputs_shape_new = (-1,) + outputs_shape_new[1:]
             return self.reshape(outputs, outputs_shape_new)
 
-        unstack = Unstack(time_axis)
+        unstack = ops.Unstack(time_axis)
         inputs = unstack(inputs)
         y = ()
         for item in inputs:
@@ -151,5 +151,5 @@ class TimeDistributed(Cell):
             _check_data(isinstance(outputs, Tensor), self.cls_name)
             _check_expand_dims_axis(time_axis, outputs.ndim, self.cls_name)
             y += (outputs,)
-        y = Stack(time_axis)(y)
+        y = ops.Stack(time_axis)(y)
         return y

mindspore/nn/layer/transformer.py

@@ -26,12 +26,12 @@ from mindspore.common.tensor import Tensor
 from mindspore.common.parameter import Parameter
 from mindspore.common.initializer import initializer, XavierNormal, XavierUniform, \
     HeUniform, Uniform, _calculate_fan_in_and_fan_out
-from mindspore.ops.function.nn_func import multi_head_attention_forward
 from mindspore.nn.cell import Cell
 from .basic import Dense, Dropout
 from .activation import ReLU, GELU
 from .normalization import LayerNorm
 from .container import CellList
+
 __all__ = ['MultiheadAttention', 'TransformerEncoderLayer', 'TransformerDecoderLayer',
            'TransformerEncoder', 'TransformerDecoder', 'Transformer']
 
@@ -212,7 +212,7 @@ class MultiheadAttention(Cell):
                 query, key, value = [x.swapaxes(1, 0) for x in (query, key, value)]
 
         if not self._qkv_same_embed_dim:
-            attn_output, attn_output_weights = multi_head_attention_forward(
+            attn_output, attn_output_weights = ops.function.nn_func.multi_head_attention_forward(
                 query, key, value, self.embed_dim, self.num_heads,
                 self.in_proj_weight, self.in_proj_bias,
                 self.bias_k, self.bias_v, self.add_zero_attn,
@@ -224,7 +224,7 @@ class MultiheadAttention(Cell):
                 v_proj_weight=self.v_proj_weight, average_attn_weights=average_attn_weights,
                 k_is_v=self.k_is_v, q_is_k=self.q_is_k, dtype=self.dtype)
         else:
-            attn_output, attn_output_weights = multi_head_attention_forward(
+            attn_output, attn_output_weights = ops.function.nn_func.multi_head_attention_forward(
                 query, key, value, self.embed_dim, self.num_heads,
                 self.in_proj_weight, self.in_proj_bias,
                 self.bias_k, self.bias_v, self.add_zero_attn,
@@ -328,7 +328,7 @@ class TransformerEncoderLayer(Cell):
         self.activation1 = activation
 
         if not isinstance(activation, str) and not isinstance(activation, Cell) \
-            and not callable(activation):
+                and not callable(activation):
             raise ValueError(f"The argument 'activation' must be str, callable or Cell instance,"
                              f" but get {activation}.")
         if isinstance(activation, Cell) and (not isinstance(activation, ReLU) and \
@@ -360,15 +360,23 @@ class TransformerEncoderLayer(Cell):
                 raise AssertionError(
                     "only bool and floating types of key_padding_mask are supported")
 
-        x = src
+        input_data = src
+
         if self.norm_first:
-            x = x + self._sa_block(self.norm1(x), src_mask, src_key_padding_mask)
-            x = x + self._ff_block(self.norm2(x))
+            normed_input = self.norm1(input_data)
+            sa_block_result = self._sa_block(normed_input, src_mask, src_key_padding_mask)
+            input_data = input_data + sa_block_result
+            normed_updated_input = self.norm2(input_data)
+            ff_block_result = self._ff_block(normed_updated_input)
+            input_data = input_data + ff_block_result
         else:
-            x = self.norm1(x + self._sa_block(x, src_mask, src_key_padding_mask))
-            x = self.norm2(x + self._ff_block(x))
+            sa_block_result = self._sa_block(input_data, src_mask, src_key_padding_mask)
+            normed_sa_result = self.norm1(input_data + sa_block_result)
+            input_data = normed_sa_result
+            ff_block_result = self._ff_block(input_data)
+            input_data = self.norm2(input_data + ff_block_result)
 
-        return x
+        return input_data
 
     def _sa_block(self, x, attn_mask, key_padding_mask):
         x = self.self_attn(x, x, x,
@@ -480,7 +488,7 @@ class TransformerDecoderLayer(Cell):
         self.activation1 = activation
 
         if not isinstance(activation, str) and not isinstance(activation, Cell) \
-            and not callable(activation):
+                and not callable(activation):
             raise ValueError(f"The argument 'activation' must be str, callable or Cell instance,"
                              f" but get {activation}.")
         if isinstance(activation, Cell) and (not isinstance(activation, ReLU) and \
@@ -507,17 +515,29 @@ class TransformerDecoderLayer(Cell):
     def construct(self, tgt: Tensor, memory: Tensor, tgt_mask: Optional[Tensor] = None,
                   memory_mask: Optional[Tensor] = None, tgt_key_padding_mask: Optional[Tensor] = None,
                   memory_key_padding_mask: Optional[Tensor] = None):
-        x = tgt
+        input_data = tgt
+
         if self.norm_first:
-            x = x + self._sa_block(self.norm1(x), tgt_mask, tgt_key_padding_mask)
-            x = x + self._mha_block(self.norm2(x), memory, memory_mask, memory_key_padding_mask)
-            x = x + self._ff_block(self.norm3(x))
+            normed_input = self.norm1(input_data)
+            sa_block_result = self._sa_block(normed_input, tgt_mask, tgt_key_padding_mask)
+            input_data = input_data + sa_block_result
+            normed_updated_input_1 = self.norm2(input_data)
+            mha_block_result = self._mha_block(normed_updated_input_1, memory, memory_mask, memory_key_padding_mask)
+            input_data = input_data + mha_block_result
+            normed_updated_input_2 = self.norm3(input_data)
+            ff_block_result = self._ff_block(normed_updated_input_2)
+            input_data = input_data + ff_block_result
         else:
-            x = self.norm1(x + self._sa_block(x, tgt_mask, tgt_key_padding_mask))
-            x = self.norm2(x + self._mha_block(x, memory, memory_mask, memory_key_padding_mask))
-            x = self.norm3(x + self._ff_block(x))
+            sa_block_result = self._sa_block(input_data, tgt_mask, tgt_key_padding_mask)
+            normed_sa_result = self.norm1(input_data + sa_block_result)
+            input_data = normed_sa_result
+            mha_block_result = self._mha_block(input_data, memory, memory_mask, memory_key_padding_mask)
+            normed_mha_result = self.norm2(input_data + mha_block_result)
+            input_data = normed_mha_result
+            ff_block_result = self._ff_block(input_data)
+            input_data = self.norm3(input_data + ff_block_result)
 
-        return x
+        return input_data
 
     def _sa_block(self, x, attn_mask, key_padding_mask):
         x = self.self_attn(x, x, x,
@@ -670,17 +690,19 @@ class TransformerDecoder(Cell):
     def construct(self, tgt: Tensor, memory: Tensor, tgt_mask: Optional[Tensor] = None,
                   memory_mask: Optional[Tensor] = None, tgt_key_padding_mask: Optional[Tensor] = None,
                   memory_key_padding_mask: Optional[Tensor] = None):
-        output = tgt
+        processed_output = tgt
         for mod in self.layers:
-            output = mod(output, memory, tgt_mask=tgt_mask,
-                         memory_mask=memory_mask,
-                         tgt_key_padding_mask=tgt_key_padding_mask,
-                         memory_key_padding_mask=memory_key_padding_mask)
+            layer_output = mod(processed_output, memory,
+                               tgt_mask=tgt_mask,
+                               memory_mask=memory_mask,
+                               tgt_key_padding_mask=tgt_key_padding_mask,
+                               memory_key_padding_mask=memory_key_padding_mask)
+            processed_output = layer_output
 
         if self.norm is not None:
-            output = self.norm(output)
+            processed_output = self.norm(processed_output)
 
-        return output
+        return processed_output
 
 
 class Transformer(Cell):

mindspore/nn/learning_rate_schedule.py

@@ -223,7 +223,8 @@ class InverseDecayLR(LearningRateSchedule):
         learning_rate (float): The initial value of learning rate.
         decay_rate (float): The decay rate.
         decay_steps (int): Number of steps to decay over.
-        is_stair (bool): If true, learning rate decay once every `decay_steps` times. Default: ``False`` .
+        is_stair (bool): If true, learning rate decay once every `decay_steps` times. If False, the learning rate
+            decays for every step. Default: ``False`` .
 
     Inputs:
         - **global_step** (Tensor) - The current step number.
@@ -454,8 +455,9 @@ class WarmUpLR(LearningRateSchedule):
         tmp\_step= \min(current\_step, warmup\_steps)
 
     Args:
-        learning_rate (float): The initial value of learning rate.
-        warmup_steps (int): The warm up steps of learning rate.
+        learning_rate (float): The initial value of learning rate. The value of `learning_rate` must be greater than 0.
+        warmup_steps (int): The warm up steps of learning rate. The value of `warmup_steps` must be greater than
+            or equal to 1.
 
     Inputs:
         - **global_step** (Tensor) - The current step number. Shape is :math:`()`.

mindspore/nn/loss/loss.py

@@ -24,8 +24,6 @@ from mindspore.common.tensor import Tensor
 from mindspore.common.parameter import Parameter
 from mindspore.ops import operations as P
 from mindspore.ops.operations import _inner_ops as inner
-from mindspore.ops.operations.nn_ops import MultiMarginLoss as MultiMarginLossOp
-from mindspore.ops.operations.nn_ops import MultilabelMarginLoss as MultilabelMarginLossOp
 from mindspore.ops import functional as F
 from mindspore import nn
 from mindspore.ops.primitive import constexpr, _primexpr
@@ -33,7 +31,6 @@ from mindspore.nn.cell import Cell
 from mindspore.nn.layer.activation import get_activation
 from mindspore import _checkparam as validator
 from mindspore import context
-from mindspore.ops.auto_generate import l1_loss_ext_op
 
 
 class LossBase(Cell):
@@ -319,7 +316,7 @@ class L1LossExt(LossBase):
         self.reduction = reduction
 
     def construct(self, logits, labels):
-        return l1_loss_ext_op(logits, labels, self.reduction)
+        return ops.auto_generate.l1_loss_ext_op(logits, labels, self.reduction)
 
 
 class MSELoss(LossBase):
@@ -628,11 +625,11 @@ class SmoothL1Loss(LossBase):
     .. math::
         L_{i} =
         \begin{cases}
-        \frac{0.5 (x_i - y_i)^{2}}{\beta}, & \text{if } |x_i - y_i| < {\beta} \\
-        |x_i - y_i| - 0.5 {\beta}, & \text{otherwise.}
+        \frac{0.5 (x_i - y_i)^{2}}{\text{beta}}, & \text{if } |x_i - y_i| < \text{beta} \\
+        |x_i - y_i| - 0.5 * {\text{beta}}, & \text{otherwise.}
         \end{cases}
 
-    Where :math:`{\beta}` represents the threshold `beta`.
+    Where :math:`{\text{beta}}` represents the threshold `beta`.
 
     If `reduction` is not `none`, then:
 
@@ -653,8 +650,11 @@ class SmoothL1Loss(LossBase):
           robust to outliers, and the loss function has better robustness.
 
     Args:
-        beta (float): The loss function calculates the threshold of the transformation between L1Loss and L2Loss.
-            Default: ``1.0`` .
+        beta (number, optional): The loss function calculates the threshold of the transformation
+            between L1Loss and L2Loss. Default: ``1.0`` .
+
+            - Ascend: The value should be equal to or greater than zero.
+            - CPU/GPU: The value should be greater than zero.
         reduction (str, optional): Apply specific reduction method to the output: ``'none'`` , ``'mean'`` ,
             ``'sum'`` . Default: ``'none'`` .
 
@@ -663,22 +663,26 @@ class SmoothL1Loss(LossBase):
             - ``'sum'``: the output elements will be summed.
 
     Inputs:
-        - **logits** (Tensor) - Predictive value. Tensor of any dimension. Data type must be one of float16 or
-          float32.
-        - **labels** (Tensor) - Ground truth data, same shape and dtype as the `logits`.
+        - **logits** (Tensor) - Predictive value. Tensor of any dimension. Supported dtypes:
 
+          - Ascend: float16, float32, bfloat16.
+          - CPU/GPU: float16, float32, float64.
+        - **labels** (Tensor) - Ground truth data.
+
+          - CPU/Ascend: has the same shape as the `logits`,
+            `logits` and `labels` comply with the implicit type conversion rules to make the data types consistent.
+          - GPU: has the same shape and dtype as the `logits`.
     Outputs:
         Tensor, if `reduction` is ``'none'``, then output is a tensor with the same shape as `logits`.
         Otherwise the shape of output tensor is :math:`()`.
 
     Raises:
-        TypeError: If `beta` is not a float.
-        ValueError: If `reduction` is not one of ``'none'``, ``'mean'``, ``'sum'``.
-        TypeError: If `logits` or `labels` are not Tensor.
-        TypeError: If dtype of `logits` or `labels` is neither float16 not float32.
-        TypeError: If dtype of `logits` is not the same as `labels`.
-        ValueError: If `beta` is less than or equal to 0.
+        TypeError: If input `logits` or `labels` are not Tensor.
+        RuntimeError: If dtype of `logits` or `labels` is not one of float16, float32, float64, bfloat16.
         ValueError: If shape of `logits` is not the same as `labels`.
+        ValueError: If `reduction` is not one of ``'none'``, ``'mean'``, ``'sum'``.
+        TypeError: If `beta` is not a float, int or bool.
+        RuntimeError: If `beta` is less than or equal to 0.
 
     Supported Platforms:
         ``Ascend`` ``GPU`` ``CPU``
@@ -1631,7 +1635,7 @@ class MultiMarginLoss(LossBase):
     def __init__(self, p=1, margin=1.0, reduction='mean', weight=None):
         """Initialize MultiMarginLoss."""
         super(MultiMarginLoss, self).__init__()
-        self.multi_margin_loss = MultiMarginLossOp(p=p, margin=margin, reduction=reduction)
+        self.multi_margin_loss = ops.MultiMarginLoss(p=p, margin=margin, reduction=reduction)
         self.weight = weight
 
     def construct(self, x, target, weight=None):
@@ -1718,22 +1722,11 @@ class BCELoss(LossBase):
     def __init__(self, weight=None, reduction='mean'):
         """Initialize BCELoss."""
         super(BCELoss, self).__init__(reduction)
-        self.binary_cross_entropy = P.BinaryCrossEntropy(reduction=reduction)
-        self.weight_one = weight is None
-        if not self.weight_one:
-            self.weight = weight
-        else:
-            self.ones = P.OnesLike()
+        self.reduction = reduction
+        self.weight = weight
 
     def construct(self, logits, labels):
-        _check_is_tensor('logits', logits, self.cls_name)
-        _check_is_tensor('labels', labels, self.cls_name)
-        if self.weight_one:
-            weight = self.ones(logits)
-        else:
-            weight = self.weight
-        loss = self.binary_cross_entropy(logits, labels, weight)
-        return loss
+        return F.binary_cross_entropy(logits, labels, self.weight, self.reduction)
 
 
 class CosineEmbeddingLoss(LossBase):
@@ -1887,7 +1880,7 @@ class MultilabelMarginLoss(LossBase):
 
     def __init__(self, reduction='mean'):
         super(MultilabelMarginLoss, self).__init__()
-        self.multilabel_margin_loss = MultilabelMarginLossOp(reduction=reduction)
+        self.multilabel_margin_loss = ops.MultilabelMarginLoss(reduction=reduction)
 
     def construct(self, x, target):
         loss, _ = self.multilabel_margin_loss(x, target)
@@ -2265,7 +2258,8 @@ class TripletMarginLoss(LossBase):
             - ``'mean'``: compute and return the mean of elements in the output.
             - ``'sum'``: the output elements will be summed.
 
-        margin (Union[Tensor, float]): Make a margin between the positive pair and the negative pair.
+        margin (Union[Tensor, float]): Make a margin between the positive pair and the negative pair. The length of
+            shape of `margin` must be 0.
             Default: ``1.0`` .
 
     Inputs:
@@ -2275,7 +2269,8 @@ class TripletMarginLoss(LossBase):
           shape as `x`. :math:`p` in the above formula.
         - **negative** (Tensor) - A sample belonging to the different class from `x`, with the same type and shape
           as `x`. :math:`n` in the above formula.
-        - **margin** (Union[Tensor, float]) - Make a margin between the positive pair and the negative pair.
+        - **margin** (Union[Tensor, float]) - Make a margin between the positive pair and the negative pair. The length
+          of shape of `margin` must be 0.
           Default: ``1.0`` .
 
     Outputs:

mindspore/nn/optim/ada_grad.py

@@ -78,6 +78,7 @@ class Adagrad(Optimizer):
     :math:`state\_sum` stands for the accumulated squared sum of the gradients :math:`accum`.
     :math:`g` stands for `grads`, :math:`\lambda` stands for `weight_decay`.
     :math:`\gamma` stands for `learning_rate`, :math:`w` stands for `params`.
+    :math:`t` represents current `step`.
 
     Note:
         If parameters are not grouped, the `weight_decay` in optimizer will be applied on the network parameters without

mindspore/nn/optim/adadelta.py

@@ -134,9 +134,9 @@ class Adadelta(Optimizer):
 
     Raises:
         TypeError: If `learning_rate` is not one of int, float, Tensor, Iterable, LearningRateSchedule.
-        TypeError: If element of `parameters` is neither Parameter nor dict.
+        TypeError: If element of `params` is neither Parameter nor dict.
         TypeError: If `rho`, `epsilon` or `loss_scale` is not a float.
-        TypeError: If `weight_decay` is neither float nor int.
+        TypeError: If `weight_decay` is not float, int or cell.
         ValueError: if `rho` is not in range [0.0, 1.0].
         ValueError: If `loss_scale` is less than or equal to 0.
         ValueError: If `learning_rate`, `epsilon` or `weight_decay` is less than 0.

mindspore/nn/optim/adam.py

@@ -633,7 +633,7 @@ class Adam(Optimizer):
     Raises:
         KeyError: If kwargs got keys other than 'use_lazy' or 'use_offload'.
         TypeError: If `learning_rate` is not one of int, float, Tensor, Iterable, LearningRateSchedule.
-        TypeError: If element of `parameters` is neither Parameter nor dict.
+        TypeError: If element of `params` is neither Parameter nor dict.
         TypeError: If `beta1`, `beta2`, `eps` or `loss_scale` is not a float.
         TypeError: If `weight_decay` is neither float nor int.
         TypeError: If `use_locking`, `use_nesterov`, `use_amsgrad`, `use_lazy` or `use_offload` is not a bool.

mindspore/nn/optim/lars.py

@@ -82,7 +82,7 @@ class LARS(Optimizer):
             &\hline \\[-1.ex]
         \end{array}
 
-    :math:`w` represents the network parameters, :math:`g` represents `gradients`,
+    :math:`w` represents the network's params, :math:`g` represents `gradients`,
     :math:`t` represents the current step, :math:`\lambda` represents `weight_decay` in `optimizer`,
     :math:`\gamma` represents `learning_rate` in `optimizer`, :math:`\eta` represents `coefficient`.
 
@@ -98,9 +98,6 @@ class LARS(Optimizer):
         - **gradients** (tuple[Tensor]) - The gradients of `params` in the optimizer, the shape is the
           as same as the `params` in the optimizer.
 
-    Outputs:
-        Union[Tensor[bool], tuple[Parameter]], it depends on the output of `optimizer`.
-
     Supported Platforms:
         ``Ascend``
 

mindspore/nn/optim/optimizer.py

@@ -848,7 +848,7 @@ class Optimizer(Cell):
             optim_result(bool): The results of updating parameters. This input is used to ensure that the parameters are
               updated before they are broadcast.
         Returns:
-             bool, the status flag.
+             The broadcast parameters.
         """
         # If rank_id is 0, 1, 2, 3, there are param0 ~ param7,
         # then the value is[(param0, param4), (param1, param5), (param2, param6), (param3, param7)]

mindspore/nn/optim/rprop.py

@@ -44,8 +44,8 @@ class Rprop(Optimizer):
             &\hspace{15mm} w_{t} \leftarrow w_{t-1}- \Delta_{t} \mathrm{sign}(g_t)                 \\
         \end{gather*}
 
-    :math:`\Delta_{min/max}` represents the min/max step size, :math:`\eta_{+/-}` represents the factors of
-    etaminus and etaplus, :math:`g` represents `gradients`, :math:`w` represents `parameters`.
+    :math:`g` represents `gradients`, :math:`w` represents `parameters`, :math:`\Delta_{min/max}` represents the
+    min/max step size, :math:`\eta_{+/-}` represents the factors of etaminus and etaplus.
 
     Note:
         If parameters are not grouped, the `weight_decay` in optimizer will be applied on the parameters without 'beta'

mindspore/nn/optim/thor.py

@@ -21,6 +21,7 @@ from mindspore.ops import functional as F, composite as C, operations as P
 from mindspore.common.initializer import initializer
 from mindspore.common.parameter import Parameter, ParameterTuple
 from mindspore.common.tensor import Tensor
+from mindspore.common import set_recursion_limit
 import mindspore.ops as ops
 import mindspore.nn as nn
 import mindspore.common.dtype as mstype
@@ -355,7 +356,7 @@ def thor(net, learning_rate, damping, momentum, weight_decay=0.0, loss_scale=1.0
         ...                                                 amp_level="O2", keep_batchnorm_fp32=False)
 
     """
-    context.set_context(max_call_depth=10000)
+    set_recursion_limit(10000)
     ConvertNetUtils().convert_to_thor_net(net)
     if context.get_context("device_target") == "Ascend":
         return ThorAscend(net, learning_rate, damping, momentum, weight_decay, loss_scale, batch_size, decay_filter,