比较与torch.broadcast_to的差异

查看源文件

以下映射关系均可参考本文。

PyTorch APIs

MindSpore APIs

torch.broadcast_to

mindspore.ops.broadcast_to

torch.Tensor.broadcast_to

mindspore.Tensor.broadcast_to

torch.broadcast_to

torch.broadcast_to(input, shape) -> Tensor

更多内容详见torch.broadcast_to

mindspore.ops.broadcast_to

mindspore.ops.broadcast_to(input, shape) -> Tensor

更多内容详见mindspore.ops.broadcast_to

差异对比

PyTorch:将输入shape广播到目标shape。

MindSpore:MindSpore此API实现功能与PyTorch基本一致,额外支持shape中存在-1维度的情况。如果目标shape中有-1维度,它被该维度中的输入shape的值替换。如果目标shape中有-1维度,则-1维度不能位于一个不存在的维度中。

分类

子类

PyTorch

MindSpore

差异

输入

单输入

input

input

功能一致

参数

参数1

shape

shape

功能一致

代码示例1

# PyTorch
import torch

shape = (2, 3)
x = torch.tensor([[1], [2]]).float()
torch_output = torch.broadcast_to(x, shape)
print(torch_output.numpy())
# [[1. 1. 1.]
#  [2. 2. 2.]]

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

shape = (2, 3)
x = Tensor(np.array([[1], [2]]).astype(np.float32))
output = mindspore.ops.function.broadcast_to(x, shape)
print(output)
# [[1. 1. 1.]
#  [2. 2. 2.]]