335 Star 1.5K Fork 859

MindSpore / docs

加入 Gitee
与超过 1200万 开发者一起发现、参与优秀开源项目,私有仓库也完全免费 :)
免费加入
克隆/下载
MSELoss.md 3.12 KB
一键复制 编辑 原始数据 按行查看 历史
luojianing 提交于 2023-07-21 15:16 . replace target=blank

比较与torch.nn.MSELoss的功能差异

查看源文件

torch.nn.MSELoss

torch.nn.MSELoss(size_average=None, reduce=None, reduction='mean')(input, target) -> Tensor

更多内容详见torch.nn.MSELoss

mindspore.nn.MSELoss

class mindspore.nn.MSELoss(reduction='mean')(logits, labels) -> Tensor

更多内容详见mindspore.nn.MSELoss

差异对比

PyTorch:用于计算输入input和target每一个元素的均方误差,reduction参数指定应用于loss的规约类型。

MindSpore:实现与PyTorch一致的功能。

分类 子类 PyTorch MindSpore 差异
参数 参数1 size_average - 已弃用,被reduction替代
参数2 reduce - 已弃用,被reduction替代
参数3 reduction reduction -
输入 输入1 input logits 功能一致,参数名不同
输入2 target labels 功能一致,参数名不同

代码示例1

计算inputtarget的均方误差。默认情况下,reduction='mean'

# PyTorch
import torch
from torch import nn
from torch import tensor
import numpy as np

loss = nn.MSELoss()
input_ = np.array([1, 1, 1, 1]).reshape((2, 2))
inputs = tensor(input_, dtype=torch.float32)
target_ = np.array([1, 2, 2, 1]).reshape((2, 2))
target = tensor(target_, dtype=torch.float32)
output = loss(inputs, target)
print(output.numpy())
# 0.5

# MindSpore
import mindspore
from mindspore import Tensor
import mindspore.nn as nn
import numpy as np

loss = nn.MSELoss()
input_ = np.array([1, 1, 1, 1]).reshape((2, 2))
inputs = Tensor(input_, dtype=mindspore.float32)
target_ = np.array([1, 2, 2, 1]).reshape((2, 2))
target = Tensor(target_, dtype=mindspore.float32)
output = loss(inputs, target)
print(output)
# 0.5

代码示例2

计算inputtarget的均方误差,以求和方式规约。

# PyTorch
import torch
from torch import nn
from torch import tensor
import numpy as np

loss = nn.MSELoss(reduction='sum')
input_ = np.array([1, 1, 1, 1]).reshape((2, 2))
inputs = tensor(input_, dtype=torch.float32)
target_ = np.array([1, 2, 2, 1]).reshape((2, 2))
target = tensor(target_, dtype=torch.float32)
output = loss(inputs, target)
print(output.numpy())
# 2.0

# MindSpore
import mindspore
from mindspore import Tensor
import mindspore.nn as nn
import numpy as np

loss = nn.MSELoss(reduction='sum')
input_ = np.array([1, 1, 1, 1]).reshape((2, 2))
inputs = Tensor(input_, dtype=mindspore.float32)
target_ = np.array([1, 2, 2, 1]).reshape((2, 2))
target = Tensor(target_, dtype=mindspore.float32)
output = loss(inputs, target)
print(output)
# 2.0
1
https://gitee.com/mindspore/docs.git
git@gitee.com:mindspore/docs.git
mindspore
docs
docs
r2.0

搜索帮助