diff --git a/docs/api/api_python/nn/mindspore.nn.CTCLoss.rst b/docs/api/api_python/nn/mindspore.nn.CTCLoss.rst index a997a383db0..298f779f2d0 100644 --- a/docs/api/api_python/nn/mindspore.nn.CTCLoss.rst +++ b/docs/api/api_python/nn/mindspore.nn.CTCLoss.rst @@ -9,7 +9,7 @@ mindspore.nn.CTCLoss 参数： - **blank** (int) - 空白标签。默认值：0。 - - **reduction** (str) - 指定输出结果的计算方式。可选值为"none"、"mean"或"sum"。默认值："mean"。 + - **reduction** (str) - 对输出应用归约方法。可选值为"none"、"mean"或"sum"。默认值："mean"。 - **zero_infinity** (bool) - 是否设置无限损失和相关梯度为零。默认值：False。 输入： diff --git a/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool2d.rst b/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool2d.rst index 8684f138f5d..1f12c9da9fe 100644 --- a/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool2d.rst +++ b/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool2d.rst @@ -3,8 +3,7 @@ mindspore.nn.FractionalMaxPool2d .. py:class:: mindspore.nn.FractionalMaxPool2d(kernel_size, output_size=None, output_ratio=None, return_indices=False, _random_samples=None) - 对多个输入平面组成的输入上应用2D分数最大池化。在 :math:`(kH_{in}, kW_{in})` 区域上应用最大池化操作，由输出shape决定随机步长。对于任何输入shape，指定输出shape为 :math:`(H, W)` 。输出特征的数量等于输入平面的数量。 - 在一个输入Tensor上应用2D fractional max pooling，可被视为组成一个2D平面。 + 在输入 `input_x` 上应用二维分数最大池化。输出Tensor的shape可以由 `output_size` 和 `output_ratio` 其中之一确定，步长由 `_random_samples` 决定。 `output_size` 和 `output_ratio` 不能同时使用。 分数最大池化的详细描述在 `Fractional Max-Pooling `_ 。 @@ -12,7 +11,7 @@ mindspore.nn.FractionalMaxPool2d - **kernel_size** (Union[int, tuple[int]]) - 指定池化核尺寸大小，如果为int，则代表池化核的高和宽。如果为tuple，其值必须包含两个正整数值分别表示池化核的高和宽。取值必须为正整数。 - **output_size** (Union[int, tuple[int]]，可选) - 目标输出shape。如果是整数，则表示输出目标的高和宽。如果是tuple，其值必须包含两个整数值分别表示目标输出的高和宽。默认值：None。 - **output_ratio** (Union[float, tuple[float]]，可选) - 目标输出shape与输入shape的比率。通过输入shape和 `output_ratio` 确定输出shape。支持数据类型：float16、float32、double，数值介于0到1之间。默认值：None。 - - **return_indices** (bool，可选) - 如果为 `True` ，返回分数最大池化的最大值的的索引值。默认值：False。 + - **return_indices** (bool，可选) - 是否返回最大值的的索引值。默认值：False。 - **_random_samples** (Tensor，可选) - 3D张量，分数最大池化的随机步长。支持的数据类型：float16、float32、double。数值介于0到1之间。shape为 :math:`(N, C, 2)` 的Tensor。默认值：None。 输入： diff --git a/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool3d.rst b/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool3d.rst index a60e43dd428..0f4793e2e19 100644 --- a/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool3d.rst +++ b/docs/api/api_python/nn/mindspore.nn.FractionalMaxPool3d.rst @@ -3,7 +3,7 @@ mindspore.nn.FractionalMaxPool3d .. py:class:: mindspore.nn.FractionalMaxPool3d(kernel_size, output_size=None, output_ratio=None, return_indices=False, _random_samples=None) - 对多个输入平面组成的输入上应用3D分数最大池化。在 :math:`(kD_{in}, kH_{in}, kW_{in})` 区域上应用最大池化操作，由输出shape决定随机步长。输出特征的数量等于输入平面的数量。 + 在输入 `input_x` 上应用三维分数最大池化。输出Tensor的shape可以由 `output_size` 和 `output_ratio` 其中之一确定，步长由 `_random_samples` 决定。 `output_size` 和 `output_ratio` 不能同时使用。 分数最大池化的详细描述在 `Fractional MaxPooling by Ben Graham `_ 。 @@ -13,7 +13,7 @@ mindspore.nn.FractionalMaxPool3d - **kernel_size** (Union[int, tuple[int]]) - 指定池化核尺寸大小，如果为int，则代表池化核的深度，高和宽。如果为tuple，其值必须包含三个正整数值分别表示池化核的深度，高和宽。取值必须为正整数。 - **output_size** (Union[int, tuple[int]]，可选) - 目标输出大小。如果是整数，则表示输出目标的深、高和宽。如果是tuple，其值必须包含三个整数值分别表示目标输出的深、高和宽。默认值：None。 - **output_ratio** (Union[float, tuple[float]]，可选) - 目标输出shape与输入shape的比率。通过输入shape和 `output_ratio` 确定输出shape。支持数据类型：float16、float32、double，数值介于0到1之间。默认值：None。 - - **return_indices** (bool，可选) - 如果为 `True` ，返回分数最大池化的最大值的的索引值。默认值：False。 + - **return_indices** (bool，可选) - 是否返回最大值的的索引值。默认值：False。 - **_random_samples** (Tensor，可选) - 随机步长。支持的数据类型：float16、float32、double。shape为 :math:`(N, C, 3)` 的Tensor。数值介于0到1之间。默认值：None。 输入： diff --git a/docs/api/api_python/nn/mindspore.nn.PixelShuffle.rst b/docs/api/api_python/nn/mindspore.nn.PixelShuffle.rst index 675637d3fa6..f562166596e 100644 --- a/docs/api/api_python/nn/mindspore.nn.PixelShuffle.rst +++ b/docs/api/api_python/nn/mindspore.nn.PixelShuffle.rst @@ -3,12 +3,15 @@ mindspore.nn.PixelShuffle .. py:class:: mindspore.nn.PixelShuffle(upscale_factor) - 在多个输入平面组成的输入上面应用PixelShuffle算法。在平面上应用高效亚像素卷积，步长为 :math:`1/r` 。关于PixelShuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 + 对 `x` 应用像素重组操作，它实现了步长为 :math:`1/r` 的子像素卷积。关于PixelShuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 通常情况下，输入shape :math:`(*, C \times r^2, H, W)` ，输出shape :math:`(*, C, H \times r, W \times r)` 。`r` 是缩小因子。 `*` 是大于等于0的维度。 + .. note:: + Ascend上输入Tensor的维度要小于7。 + 参数： - - **upscale_factor** (int) - 增加空间分辨率的因子，是正整数。 + - **upscale_factor** (int) - 打乱输入Tensor的因子，是正整数。 `upscale_factor` 是上面提到的 :math:`r` 。 输入： - **x** (Tensor) - Tensor，shape为 :math:`(*, C \times r^2, H, W)` 。输入Tensor的维度需要大于2，并且倒数第三维length可以被 `upscale_factor` 的平方整除。 diff --git a/docs/api/api_python/nn/mindspore.nn.PixelUnshuffle.rst b/docs/api/api_python/nn/mindspore.nn.PixelUnshuffle.rst index 1a85cbe129f..1f3340b8cd5 100644 --- a/docs/api/api_python/nn/mindspore.nn.PixelUnshuffle.rst +++ b/docs/api/api_python/nn/mindspore.nn.PixelUnshuffle.rst @@ -3,12 +3,12 @@ mindspore.nn.PixelUnshuffle .. py:class:: mindspore.nn.PixelUnshuffle(downscale_factor) - 在多个输入平面组成的输入上面应用PixelUnshuffle算法。关于PixelUnshuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 + 对 `x` 应用逆像素重组操作，这是像素重组的逆操作。关于PixelUnshuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 通常情况下，输入shape :math:`(*, C, H \times r, W \times r)` ，输出shape :math:`(*, C \times r^2, H, W)` 。`r` 是缩小因子。 `*` 是大于等于0的维度。 参数： - - **downscale_factor** (int) - 减小空间分辨率的因子，是正整数。 + - **downscale_factor** (int) - 恢复输入Tensor的因子，是正整数。 `downscale_factor` 是上面提到的 :math:`r` 。 输入： - **x** (Tensor) - Tensor，shape为 :math:`(*, C, H \times r, W \times r)` 。输入Tensor的维度需要大于2，并且倒数第一和倒数第二维length可以被 `downscale_factor` 整除。 diff --git a/docs/api/api_python/ops/mindspore.ops.func_diagflat.rst b/docs/api/api_python/ops/mindspore.ops.func_diagflat.rst index 6c1fb31141e..52028c02c2a 100644 --- a/docs/api/api_python/ops/mindspore.ops.func_diagflat.rst +++ b/docs/api/api_python/ops/mindspore.ops.func_diagflat.rst @@ -3,18 +3,18 @@ mindspore.ops.diagflat .. py:function:: mindspore.ops.diagflat(x, offset=0) - 创建一个二维Tensor，用展开后的输入作为它的对角线。 + 创建一个二维Tensor，用展开后的 `x` 作为它的对角线。 参数： - **x** (Tensor) - 输入Tensor，展开后作为输出的对角线。 - **offset** (int, 可选) - `offset` 控制选择哪条对角线。默认值：0。 - 当 `offset` 是0时，选择的对角线是主对角线。 - - 当 `offset` 大于0时，选择的对角线在主对角线上。 - - 当 `offset` 小于0时，选择的对角线在主对角线下。 + - 当 `offset` 是正整数时，选择的对角线在主对角线上。 + - 当 `offset` 是负整数时，选择的对角线在主对角线下。 返回： - 二维Tensor。 + 二维Tensor，对角线是展开的 `x` 。 异常： - **TypeError** - `x` 不是Tensor。 diff --git a/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool2d.rst b/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool2d.rst index 86e7cd50609..ec537894740 100644 --- a/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool2d.rst +++ b/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool2d.rst @@ -3,8 +3,7 @@ mindspore.ops.fractional_max_pool2d .. py:function:: mindspore.ops.fractional_max_pool2d(input_x, kernel_size, output_size=None, output_ratio=None, return_indices=False, _random_samples=None) - 对多个输入平面组成的输入上应用2D分数最大池化。在 :math:`(kH_{in}, kW_{in})` 区域上应用最大池化操作，由输出shape决定随机步长。对于任何输入shape，指定输出shape为 :math:`(H, W)` 。输出特征的数量等于输入平面的数量。 - 在一个输入Tensor上应用2D fractional max pooling，可被视为组成一个2D平面。 + 在输入 `input_x` 上应用二维分数最大池化。输出Tensor的shape可以由 `output_size` 和 `output_ratio` 其中之一确定，步长由 `_random_samples` 随机。 `output_size` 和 `output_ratio` 不能同时使用。 分数最大池化的详细描述在 `Fractional Max-Pooling `_ 。 @@ -13,7 +12,7 @@ mindspore.ops.fractional_max_pool2d - **kernel_size** (Union[int, tuple[int]]) - 指定池化核尺寸大小，如果为int，则代表池化核的高和宽。如果为tuple，其值必须包含两个正int值分别表示池化核的高和宽。取值必须为正int。 - **output_size** (Union[int, tuple[int]]，可选) - 目标输出shape。如果是int，则表示输出目标的高和宽。如果是tuple，其值必须包含两个int值分别表示目标输出的高和宽。默认值：None。 - **output_ratio** (Union[float, tuple[float]]，可选) - 目标输出shape与输入shape的比率。通过输入shape和 `output_ratio` 确定输出shape。支持数据类型：float16、float32、double，数值范围（0，1）。默认值：None。 - - **return_indices** (bool，可选) - 如果为 `True` ，返回分数最大池化的最大值的的索引值。默认值：False。 + - **return_indices** (bool，可选) - 是否返回最大值的的索引值。默认值：False。 - **_random_samples** (Tensor，可选) - 3D Tensor，分数最大池化的随机步长。支持的数据类型：float16、float32、double。数值范围（0，1）。shape为 :math:`(N, C, 2)` 的Tensor。默认值：None。 返回： diff --git a/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool3d.rst b/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool3d.rst index 1d833135d11..3b782e5e1bc 100644 --- a/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool3d.rst +++ b/docs/api/api_python/ops/mindspore.ops.func_fractional_max_pool3d.rst @@ -3,7 +3,7 @@ mindspore.ops.fractional_max_pool3d .. py:function:: mindspore.ops.fractional_max_pool3d(input_x, kernel_size, output_size=None, output_ratio=None, return_indices=False, _random_samples=None) - 对多个输入平面组成的输入上应用3D分数最大池化。在 :math:`(kD_{in}, kH_{in}, kW_{in})` 区域上应用最大池化操作，由输出shape决定随机步长。输出特征的数量等于输入平面的数量。 + 在输入 `input_x` 上应用三维分数最大池化。输出Tensor的shape可以由 `output_size` 和 `output_ratio` 其中之一确定，步长由 `_random_samples` 随机。 `output_size` 和 `output_ratio` 不能同时使用。 分数最大池化的详细描述在 `Fractional MaxPooling by Ben Graham `_ 。 @@ -14,7 +14,7 @@ mindspore.ops.fractional_max_pool3d - **kernel_size** (Union[int, tuple[int]]) - 指定池化核尺寸大小，如果为int，则代表池化核的深度，高和宽。如果为tuple，其值必须包含三个正int值分别表示池化核的深度，高和宽。取值必须为正int。 - **output_size** (Union[int, tuple[int]]，可选) - 目标输出大小。如果是int，则表示输出目标的深、高和宽。如果是tuple，其值必须包含三个int值分别表示目标输出的深、高和宽。默认值：None。 - **output_ratio** (Union[float, tuple[float]]，可选) - 目标输出shape与输入shape的比率。通过输入shape和 `output_ratio` 确定输出shape。支持数据类型：float16、float32、double，数值范围（0，1）。默认值：None。 - - **return_indices** (bool，可选) - 如果为 `True` ，返回分数最大池化的最大值的的索引值。默认值：False。 + - **return_indices** (bool，可选) - 是否返回最大值的的索引值。默认值：False。 - **_random_samples** (Tensor，可选) - 随机步长。支持的数据类型：float16、float32、double。shape为 :math:`(N, C, 3)` 的Tensor。数值范围（0，1）。默认值：None。 返回： diff --git a/docs/api/api_python/ops/mindspore.ops.func_pixel_shuffle.rst b/docs/api/api_python/ops/mindspore.ops.func_pixel_shuffle.rst index 9304a9b6a45..1bf465539e1 100644 --- a/docs/api/api_python/ops/mindspore.ops.func_pixel_shuffle.rst +++ b/docs/api/api_python/ops/mindspore.ops.func_pixel_shuffle.rst @@ -3,13 +3,16 @@ mindspore.ops.pixel_shuffle .. py:function:: mindspore.ops.pixel_shuffle(x, upscale_factor) - 在多个输入平面组成的输入上面应用pixel_shuffle算法。在平面上应用高效亚像素卷积，步长为 :math:`1/r` 。关于pixel_shuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 + 对输入 `x` 应用像素重组操作，它实现了步长为 :math:`1/r` 的子像素卷积。关于pixel_shuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 通常情况下，`x` shape :math:`(*, C \times r^2, H, W)` ，输出shape :math:`(*, C, H \times r, W \times r)` 。`r` 是缩小因子。 `*` 是大于等于0的维度。 + .. note:: + Ascend上输入Tensor的维度要小于7。 + 参数： - **x** (Tensor) - Tensor，shape为 :math:`(*, C \times r^2, H, W)` 。 `x` 的维度需要大于2，并且倒数第三维length可以被 `upscale_factor` 的平方整除。 - - **upscale_factor** (int) - 增加空间分辨率的因子，是正整数。 + - **upscale_factor** (int) - 打乱输入Tensor的因子，是正整数。 `upscale_factor` 是上面提到的 :math:`r` 。 返回： - **output** (Tensor) - Tensor，shape为 :math:`(*, C, H \times r, W \times r)` 。 diff --git a/docs/api/api_python/ops/mindspore.ops.func_pixel_unshuffle.rst b/docs/api/api_python/ops/mindspore.ops.func_pixel_unshuffle.rst index 48ee38e5470..5da36065216 100644 --- a/docs/api/api_python/ops/mindspore.ops.func_pixel_unshuffle.rst +++ b/docs/api/api_python/ops/mindspore.ops.func_pixel_unshuffle.rst @@ -3,13 +3,13 @@ mindspore.ops.pixel_unshuffle .. py:function:: mindspore.ops.pixel_unshuffle(x, downscale_factor) - 在多个输入平面组成的输入上面应用pixel_unshuffle算法。关于pixel_unshuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 + 对输入 `x` 应用逆像素重组操作，这是像素重组的逆操作。关于pixel_unshuffle算法详细介绍，请参考 `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ 。 通常情况下，`x` shape :math:`(*, C, H \times r, W \times r)` ，输出shape :math:`(*, C \times r^2, H, W)` 。`r` 是缩小因子。 `*` 是大于等于0的维度。 参数： - **x** (Tensor) - Tensor，shape为 :math:`(*, C, H \times r, W \times r)` 。 `x` 的维度需要大于2，并且倒数第一和倒数第二维length可以被 `downscale_factor` 整除。 - - **downscale_factor** (int) - 减小空间分辨率的因子，是正整数。 + - **downscale_factor** (int) - 恢复输入Tensor的因子，是正整数。 `downscale_factor` 是上面提到的 :math:`r` 。 返回： - **output** (Tensor) - Tensor，shape为 :math:`(*, C \times r^2, H, W)` 。 diff --git a/docs/api/api_python/ops/mindspore.ops.func_threshold.rst b/docs/api/api_python/ops/mindspore.ops.func_threshold.rst index b1616e92fcc..6bb46f3a733 100644 --- a/docs/api/api_python/ops/mindspore.ops.func_threshold.rst +++ b/docs/api/api_python/ops/mindspore.ops.func_threshold.rst @@ -3,7 +3,7 @@ mindspore.ops.threshold .. py:function:: mindspore.ops.threshold(input_x, thr, value) - threshold激活函数，按元素计算输出。 + 将使用 `thr` 参数对 `input_x` 逐元素阈值化后的结果作为Tensor返回。 threshold定义为： diff --git a/mindspore/python/mindspore/nn/layer/image.py b/mindspore/python/mindspore/nn/layer/image.py index 4fefc7a957c..c6e3a8bd515 100644 --- a/mindspore/python/mindspore/nn/layer/image.py +++ b/mindspore/python/mindspore/nn/layer/image.py @@ -532,16 +532,20 @@ class CentralCrop(Cell): class PixelShuffle(Cell): r""" - Applies a pixelshuffle operation over an input signal composed of several input planes. This is useful for - implementiong efficient sub-pixel convolution with a stride of :math:`1/r`. For more details, refer to + Applies the PixelShuffle operation over input `x` which implements sub-pixel convolutions + with stride :math:`1/r` . For more details, refer to `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ . Typically, the input is of shape :math:`(*, C \times r^2, H, W)` , and the output is of shape :math:`(*, C, H \times r, W \times r)`, where r is an upscale factor and * is zero or more batch dimensions. + Note: + The dimension of input Tensor on Ascend should be less than 7. + Args: - upscale_factor (int): factor to increase spatial resolution by, and is a positive integer. + upscale_factor (int): factor to shuffle the input, and is a positive integer. + `upscale_factor` is the above-mentioned :math:`r`. Inputs: - **x** (Tensor) - Tensor of shape :math:`(*, C \times r^2, H, W)` . The dimension of `x` is larger than 2, and @@ -559,12 +563,12 @@ class PixelShuffle(Cell): ``Ascend`` ``GPU`` ``CPU`` Examples: - >>> input_x = np.arange(3 * 2 * 9 * 4 * 4).reshape((3, 2, 9, 4, 4)) + >>> input_x = np.arange(3 * 2 * 8 * 4 * 4).reshape((3, 2, 8, 4, 4)) >>> input_x = mindspore.Tensor(input_x, mindspore.dtype.int32) - >>> pixel_shuffle = nn.PixelShuffle(3) + >>> pixel_shuffle = nn.PixelShuffle(2) >>> output = pixel_shuffle(input_x) >>> print(output.shape) - (3, 2, 1, 12, 12) + (3, 2, 2, 8, 8) """ def __init__(self, upscale_factor): super(PixelShuffle, self).__init__() @@ -576,15 +580,16 @@ class PixelShuffle(Cell): class PixelUnshuffle(Cell): r""" - Applies a pixelunshuffle operation over an input signal composed of several input planes. For more details, refer to - `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network + Applies the PixelUnshuffle operation over input `x` which is the inverse of PixelShuffle. For more details, refer + to `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ . Typically, the input is of shape :math:`(*, C, H \times r, W \times r)` , and the output is of shape :math:`(*, C \times r^2, H, W)` , where r is a downscale factor and * is zero or more batch dimensions. Args: - downscale_factor (int): factor to decrease spatial resolution by, and is a positive integer. + downscale_factor (int): factor to unshuffle the input, and is a positive integer. + `downscale_factor` is the above-mentioned :math:`r`. Inputs: - **x** (Tensor) - Tensor of shape :math:`(*, C, H \times r, W \times r)` . The dimension of `x` is larger than @@ -602,12 +607,12 @@ class PixelUnshuffle(Cell): ``Ascend`` ``GPU`` ``CPU`` Examples: - >>> pixel_unshuffle = nn.PixelUnshuffle(3) - >>> input_x = np.arange(12 * 12).reshape((1, 1, 12, 12)) + >>> pixel_unshuffle = nn.PixelUnshuffle(2) + >>> input_x = np.arange(8 * 8).reshape((1, 1, 8, 8)) >>> input_x = mindspore.Tensor(input_x, mindspore.dtype.int32) >>> output = pixel_unshuffle(input_x) >>> print(output.shape) - (1, 9, 4, 4) + (1, 4, 4, 4) """ def __init__(self, downscale_factor): super(PixelUnshuffle, self).__init__() diff --git a/mindspore/python/mindspore/nn/layer/pooling.py b/mindspore/python/mindspore/nn/layer/pooling.py index acda25a9f03..2afdb733065 100644 --- a/mindspore/python/mindspore/nn/layer/pooling.py +++ b/mindspore/python/mindspore/nn/layer/pooling.py @@ -1190,12 +1190,11 @@ class AdaptiveMaxPool3d(Cell): class FractionalMaxPool2d(Cell): r""" - Applies a 2D fractional max pooling to an input signal composed of multiple input planes. - The max-pooling operation is applied in kH × kW regions by a stochastic step size determined by - the target output size. For any input size, the size of the specified output is H x W. The number - of output features is equal to the number of input planes. + Applies the 2D FractionalMaxPool operatin over `input_x`. The output Tensor shape can be determined by either + `output_size` or `output_ratio`, and the step size is determined by `_random_samples`. + `output_size` or `output_ratio` cannot be used at the same time. - Fractional MaxPooling is described in the paper `Fractional Max-Pooling `_. + Refer to the paper `Fractional MaxPooling by Ben Graham `_ for more details. Args: kernel_size (Union[int, tuple[int]]): The size of kernel used to take the maximum value, @@ -1211,8 +1210,7 @@ class FractionalMaxPool2d(Cell): Specifying the size of the output tensor by using a ratio of the input size. Data type : float16, float32, double, and value is between (0, 1). Default: None. - return_indices (bool, optional): If `return_indices` is True, the indices of max value would be output. - Default: False. + return_indices (bool, optional): Whether to return the indices of max value. Default: False. _random_samples (Tensor, optional): The random step of FractionalMaxPool2d, which is a 3D tensor. Tensor of data type : float16, float32, double, and value is between (0, 1). Supported shape :math:`(N, C, 2)`. @@ -1293,9 +1291,9 @@ class FractionalMaxPool2d(Cell): class FractionalMaxPool3d(Cell): r""" - This operator applies a 3D fractional max pooling over an input signal composed of several input planes. - The max-pooling operation is applied in kD x kH x kW regions by a stochastic step size determined - by the target output size.The number of output features is equal to the number of input planes. + Applies the 3D FractionalMaxPool operatin over `input_x`. The output Tensor shape can be determined by either + `output_size` or `output_ratio`, and the step size is determined by `_random_samples`. + `output_size` or `output_ratio` cannot be used at the same time. Refer to the paper `Fractional MaxPooling by Ben Graham `_ for more details. @@ -1316,8 +1314,7 @@ class FractionalMaxPool3d(Cell): Specifying the size of the output tensor by using a ratio of the input size. Data type : float16, float32, double, and value is between (0, 1). Default: None. - return_indices (bool, optional): If `return_indices` is True, the indices of max value would be output. - Default: False. + return_indices (bool, optional): Whether to return the indices of max value. Default: False. _random_samples (Tensor, optional): The random step of FractionalMaxPool3d, which is a 3D tensor. Tensor of data type : float16, float32, double, and value is between (0, 1). Supported shape :math:`(N, C, 3)` diff --git a/mindspore/python/mindspore/nn/loss/loss.py b/mindspore/python/mindspore/nn/loss/loss.py index 9b28fda8bdb..aeb00d3f79e 100644 --- a/mindspore/python/mindspore/nn/loss/loss.py +++ b/mindspore/python/mindspore/nn/loss/loss.py @@ -2302,13 +2302,13 @@ class CTCLoss(LossBase): Recurrent Neural Networks `_ . Args: - blank (int): The blank label. Default: 0. - reduction (str): Apply specific reduction method to the output: 'none', 'mean', or 'sum'. Default: 'mean'. - zero_infinity (bool): Whether to set infinite loss and correlation gradient to zero. Default: False. + blank (int): The blank tag. Default: 0. + reduction (str): Implements the reduction method to the output with 'none', 'mean', or 'sum'. Default: 'mean'. + zero_infinity (bool): Whether to set infinite loss and correlation gradient to 0. Default: False. Inputs: - - **log_probs** (Tensor) - A tensor of shape (T, N, C) or (T, C), where T is input length, N is batch size and - C is number of classes (including blank). T, N and C are positive integers. + - **log_probs** (Tensor) - A tensor of shape (T, N, C) or (T, C), where T is length of input, + N is size of the batch and C is the number of classes. T, N and C are positive integers. - **targets** (Tensor) - A tensor of shape (N, S) or (sum( `target_lengths` )), where S is max target length, means the target sequences. - **input_lengths** (Union[tuple, Tensor]) - A tuple or Tensor of shape(N). It means the lengths of the input. diff --git a/mindspore/python/mindspore/ops/function/array_func.py b/mindspore/python/mindspore/ops/function/array_func.py index b540cdea900..93895994737 100644 --- a/mindspore/python/mindspore/ops/function/array_func.py +++ b/mindspore/python/mindspore/ops/function/array_func.py @@ -502,7 +502,7 @@ def ravel(x): >>> x = Tensor(np.array([[0, 1], [2, 1]]).astype(np.float32)) >>> output = ops.ravel(x) >>> print(output) - [0. 1. 2. 1] + [0. 1. 2. 1.] >>> print(output.shape) (4,) """ @@ -4887,18 +4887,18 @@ def diag(input_x): def diagflat(x, offset=0): r""" - Creates a two-dimensional Tensor with the flattened input as a diagonal. + Create a 2-D Tensor which diagonal is the flattened `x` . Args: x (Tensor): Input Tensor, which is flattened and set as the diagonal of the output. - offset (int, optional): `offset` controls which diagonal to consider. Default: 0. + offset (int, optional): `offset` controls which diagonal to choose. Default: 0. - When `offset` is zero, the diagonal chosen is the main diagonal. - - When `offset` is greater than zero, the diagonal chosen is above the main diagonal. - - When `offset` is less than zero, the diagonal chosen is below the main diagonal. + - When `offset` is a positive integer, the diagonal chosen is up the main diagonal. + - When `offset` is a negative integer, the diagonal chosen is down the main diagonal. Returns: - The 2-D Tensor. + The 2-D Tensor, whose diagonal is the flattened `x`. Raises: TypeError: If `x` is not a tensor. diff --git a/mindspore/python/mindspore/ops/function/math_func.py b/mindspore/python/mindspore/ops/function/math_func.py index bcc2b9b693e..22eb1fd0501 100644 --- a/mindspore/python/mindspore/ops/function/math_func.py +++ b/mindspore/python/mindspore/ops/function/math_func.py @@ -9717,9 +9717,10 @@ def isposinf(x): ``Ascend`` ``GPU`` ``CPU`` Examples: - >>> output = ops.isposinf(Tensor([-float("inf"), float("inf"), 1.2], mstype.float32)) + >>> output = ops.isposinf(Tensor([[-float("inf"), float("inf")], [1, float("inf")]], mstype.float32)) >>> print(output) - [False True False] + [[False True] + [False True]] """ if not isinstance(x, (Tensor, Tensor_)): raise TypeError(f"For isposinf, the input x must be a Tensor, but got {type(x)}") @@ -9743,9 +9744,10 @@ def isneginf(x): ``Ascend`` ``GPU`` ``CPU`` Examples: - >>> output = ops.isneginf(Tensor([-float("inf"), float("inf"), 1.2], mstype.float32)) + >>> output = ops.isneginf(Tensor([[-float("inf"), float("inf")], [1, -float("inf")]], mstype.float32)) >>> print(output) - [ True False False] + [[ True False] + [False True]] """ if not isinstance(x, (Tensor, Tensor_)): raise TypeError(f"For isneginf, the input x must be a Tensor, but got {type(x)}") diff --git a/mindspore/python/mindspore/ops/function/nn_func.py b/mindspore/python/mindspore/ops/function/nn_func.py index d3adfc3d46f..58800ebb315 100644 --- a/mindspore/python/mindspore/ops/function/nn_func.py +++ b/mindspore/python/mindspore/ops/function/nn_func.py @@ -1455,13 +1455,11 @@ def _check_float_range_inc_right(arg_value, lower_limit, upper_limit, arg_name=N def fractional_max_pool2d(input_x, kernel_size, output_size=None, output_ratio=None, return_indices=False, _random_samples=None): r""" - Applies a 2D fractional max pooling to an input signal. - The input is composed of multiple input planes. - The max-pooling operation is applied in kH × kW regions by a stochastic step size determined by - the target output size. For any input size, the size of the specified output is H x W. The number - of output features is equal to the number of input planes. + Applies the 2D FractionalMaxPool operatin over `input_x`. The output Tensor shape can be determined by either + `output_size` or `output_ratio`, and the step size is determined by `_random_samples`. + `output_size` or `output_ratio` cannot be used at the same time. - Fractional MaxPooling is described in the paper `Fractional Max-Pooling `_. + Refer to the paper `Fractional MaxPooling by Ben Graham `_ for more details. Args: input_x (Tensor): Tensor of shape :math:`(N, C, H_{in}, W_{in})`, @@ -1479,8 +1477,7 @@ def fractional_max_pool2d(input_x, kernel_size, output_size=None, output_ratio=N Specifying the size of the output tensor by using a ratio of the input size. Data type: float16, float32, double, and value is between (0, 1). Default: None. - return_indices (bool, optional): If `return_indices` is True, the indices of max value would be output. - Default: False. + return_indices (bool, optional): Whether to return the indices of max value. Default: False. _random_samples (Tensor, optional): The random step of FractionalMaxPool2d, which is a 3D tensor. Tensor of data type: float16, float32, double, and value is between (0, 1). Supported shape :math:`(N, C, 2)`. @@ -1556,10 +1553,9 @@ def fractional_max_pool2d(input_x, kernel_size, output_size=None, output_ratio=N def fractional_max_pool3d(input_x, kernel_size, output_size=None, output_ratio=None, return_indices=False, _random_samples=None): r""" - This operator applies a 3D fractional max pooling over an input signal. - The input is composed of several input planes. - The max-pooling operation is applied in kD x kH x kW regions by a stochastic step size determined - by the target output size.The number of output features is equal to the number of input planes. + Applies the 3D FractionalMaxPool operatin over `input_x`. The output Tensor shape can be determined by either + `output_size` or `output_ratio`, and the step size is determined by `_random_samples`. + `output_size` or `output_ratio` cannot be used at the same time. Refer to the paper `Fractional MaxPooling by Ben Graham `_ for more details. @@ -1583,8 +1579,7 @@ def fractional_max_pool3d(input_x, kernel_size, output_size=None, output_ratio=N Specifying the size of the output tensor by using a ratio of the input size. Data type: float16, float32, double, and value is between (0, 1). Default: None. - return_indices (bool, optional): If `return_indices` is True, the indices of max value would be output. - Default: False. + return_indices (bool, optional): Whether to return the indices of max value. Default: False. _random_samples (Tensor, optional): The random step of FractionalMaxPool3d, which is a 3D tensor. Tensor of data type: float16, float32, double, and value is between (0, 1). Supported shape :math:`(N, C, 3)`. @@ -3414,7 +3409,7 @@ def smooth_l1_loss(logits, labels, beta=1.0, reduction='none'): def threshold(input_x, thr, value): r""" - thresholds each element of the input Tensor. + Returns each element of `input_x` after thresholding by `thr` as a Tensor. The formula is defined as follows: @@ -3427,7 +3422,7 @@ def threshold(input_x, thr, value): Args: input_x (Tensor): The input of threshold with data type of float16 or float32. - thr (Union[int, float]): The value to threshold at. + thr (Union[int, float]): The value of the threshold. value (Union[int, float]): The value to replace with when element is less than threshold. Returns: @@ -3442,10 +3437,10 @@ def threshold(input_x, thr, value): ``Ascend`` ``GPU`` ``CPU`` Examples: - >>> inputs = mindspore.Tensor([0.0, 0.2, 0.3], mindspore.float32) - >>> outputs = ops.threshold(inputs, 0.1, 20) + >>> inputs = mindspore.Tensor([0.0, 2, 3], mindspore.float32) + >>> outputs = ops.threshold(inputs, 1, 100) >>> print(outputs) - [ 20.0 0.2 0.3] + [100. 2. 3.] """ _check_is_tensor('input_x', input_x, "threshold") _check_value_type("thr", thr, [float, int], "threshold") @@ -4984,18 +4979,22 @@ def _check_positive_int(arg_value, arg_name=None, prim_name=None): def pixel_shuffle(x, upscale_factor): r""" - Applies a pixel_shuffle operation over an input signal composed of several input planes. This is useful for - implementiong efficient sub-pixel convolution with a stride of :math:`1/r`. For more details, refer to + Applies the PixelShuffle operation over input `x` which implements sub-pixel convolutions + with stride :math:`1/r` . For more details, refer to `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ . Typically, the `x` is of shape :math:`(*, C \times r^2, H, W)` , and the output is of shape :math:`(*, C, H \times r, W \times r)`, where `r` is an upscale factor and `*` is zero or more batch dimensions. + Note: + The dimension of input Tensor on Ascend should be less than 7. + Args: x (Tensor): Tensor of shape :math:`(*, C \times r^2, H, W)` . The dimension of `x` is larger than 2, and the length of third to last dimension can be divisible by `upscale_factor` squared. - upscale_factor (int): factor to increase spatial resolution by, and is a positive integer. + upscale_factor (int): factor to shuffle the input Tensor, and is a positive integer. + `upscale_factor` is the above-mentioned :math:`r`. Returns: - **output** (Tensor) - Tensor of shape :math:`(*, C, H \times r, W \times r)` . @@ -5040,7 +5039,7 @@ def pixel_shuffle(x, upscale_factor): def pixel_unshuffle(x, downscale_factor): r""" - Applies a pixel_unshuffle operation over an input signal composed of several input planes. For more details, refer + Applies the PixelUnshuffle operation over input `x` which is the inverse of PixelShuffle. For more details, refer to `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network `_ . @@ -5050,7 +5049,8 @@ def pixel_unshuffle(x, downscale_factor): Args: x (Tensor): Tensor of shape :math:`(*, C, H \times r, W \times r)` . The dimension of `x` is larger than 2, and the length of second to last dimension or last dimension can be divisible by `downscale_factor` . - downscale_factor (int): factor to decrease spatial resolution by, and is a positive integer. + downscale_factor (int): factor to unshuffle the input Tensor, and is a positive integer. + `downscale_factor` is the above-mentioned :math:`r`. Returns: - **output** (Tensor) - Tensor of shape :math:`(*, C \times r^2, H, W)` . @@ -5064,11 +5064,11 @@ def pixel_unshuffle(x, downscale_factor): ``Ascend`` ``GPU`` ``CPU`` Examples: - >>> input_x = np.arange(12 * 12).reshape((1, 1, 12, 12)) + >>> input_x = np.arange(8 * 8).reshape((1, 1, 8, 8)) >>> input_x = mindspore.Tensor(input_x, mindspore.dtype.int32) - >>> output = ops.pixel_unshuffle(input_x, 3) + >>> output = ops.pixel_unshuffle(input_x, 2) >>> print(output.shape) - (1, 9, 4, 4) + (1, 4, 4, 4) """ _check_positive_int(downscale_factor, "downscale_factor") idx = x.shape