比较与torch.std_mean的功能差异

torch.std_mean

torch.std_mean(
    input,
    dim,
    unbiased=True,
    keepdim=False
)

更多内容详见torch.std_mean。

mindspore.ops.ReduceMean

class mindspore.ops.ReduceMean(keep_dims=False)(
    input_x,
    axis=()
)

更多内容详见mindspore.ops.ReduceMean。

使用方式

PyTorch：计算指定维度数据的标准差和平均值。

MindSpore：计算指定维度数据的平均值。

代码示例

from mindspore import Tensor
import mindspore.ops as ops
import torch
import numpy as np

# In MindSpore, only the mean of given dimension will be returned.
input_x = Tensor(np.array([[1, 2], [3, 4]]).astype(np.float32))
op = ops.ReduceMean(keep_dims=True)
output = op(x=input_x, axis=1)
print(output)
# Out：
# [[1.5]
#  [3.5]]

# In torch, both std and mean of given dimensions will be returned.
input_x = torch.tensor(np.array([[1, 2], [3, 4]]).astype(np.float32))
output = torch.std_mean(input=input_x, dim=1)
std, mean = output
print('std: {}'.format(std))
print('mean: {}'.format(mean))
# Out：
# torch.tensor([0.7071, 0.7071])
# torch.tensor([1.5000, 3.5000])