比较与torch.meshgrid的差异

torch.meshgrid

torch.meshgrid(
    *tensors)

更多内容详见torch.meshgrid。

mindspore.ops.meshgrid

mindspore.ops.meshgrid(*inputs, indexing='xy')

更多内容详见mindspore.ops.meshgrid。

差异对比

PyTorch：从给定的tensors生成网格矩阵。tensors如果是scalar的list，则scalar将自动被视为大小为(1,)的张量。

MindSpore：MindSpore此API实现功能与PyTorch一致。inputs参数只支持Tensor，不支持scalar。

分类	子类	PyTorch	MindSpore	差异
参数	参数1	tensors	inputs	功能一致
	参数2	-	indexing	torch.meshgrid v1.8.1无参数indexing，其功能与MindSpore indexing设置为’ij’的功能一致；从v1.10开始，torch.meshgrid支持indexing参数

代码示例1

# PyTorch
import numpy as np
import torch

x = torch.tensor(np.array([1, 2, 3, 4]).astype(np.int32))
y = torch.tensor(np.array([5, 6, 7]).astype(np.int32))
z = torch.tensor(np.array([8, 9, 0, 1, 2]).astype(np.int32))
output = torch.meshgrid(x, y, z)
print(output)
# (tensor([[[1, 1, 1, 1, 1],
#          [1, 1, 1, 1, 1],
#          [1, 1, 1, 1, 1]],
#         [[2, 2, 2, 2, 2],
#          [2, 2, 2, 2, 2],
#          [2, 2, 2, 2, 2]],
#         [[3, 3, 3, 3, 3],
#          [3, 3, 3, 3, 3],
#          [3, 3, 3, 3, 3]],
#         [[4, 4, 4, 4, 4],
#          [4, 4, 4, 4, 4],
#          [4, 4, 4, 4, 4]]], dtype=torch.int32), tensor([[[5, 5, 5, 5, 5],
#          [6, 6, 6, 6, 6],
#          [7, 7, 7, 7, 7]],
#         [[5, 5, 5, 5, 5],
#          [6, 6, 6, 6, 6],
#          [7, 7, 7, 7, 7]],
#         [[5, 5, 5, 5, 5],
#          [6, 6, 6, 6, 6],
#          [7, 7, 7, 7, 7]],
#         [[5, 5, 5, 5, 5],
#          [6, 6, 6, 6, 6],
#          [7, 7, 7, 7, 7]]], dtype=torch.int32), tensor([[[8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2]],
#         [[8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2]],
#         [[8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2]],
#         [[8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2],
#          [8, 9, 0, 1, 2]]], dtype=torch.int32))


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

x = Tensor(np.array([1, 2, 3, 4]).astype(np.int32))
y = Tensor(np.array([5, 6, 7]).astype(np.int32))
z = Tensor(np.array([8, 9, 0, 1, 2]).astype(np.int32))
output = mindspore.ops.meshgrid(x, y, z, indexing='ij')
print(output)
# (Tensor(shape=[4, 3, 5], dtype=Int32, value=
#     [[[1, 1, 1, 1, 1],
#       [1, 1, 1, 1, 1],
#       [1, 1, 1, 1, 1]],
#      [[2, 2, 2, 2, 2],
#       [2, 2, 2, 2, 2],
#       [2, 2, 2, 2, 2]],
#      [[3, 3, 3, 3, 3],
#       [3, 3, 3, 3, 3],
#       [3, 3, 3, 3, 3]],
#      [[4, 4, 4, 4, 4],
#       [4, 4, 4, 4, 4],
#       [4, 4, 4, 4, 4]]]), Tensor(shape=[4, 3, 5], dtype=Int32, value=
#     [[[5, 5, 5, 5, 5],
#       [6, 6, 6, 6, 6],
#       [7, 7, 7, 7, 7]],
#      [[5, 5, 5, 5, 5],
#       [6, 6, 6, 6, 6],
#       [7, 7, 7, 7, 7]],
#      [[5, 5, 5, 5, 5],
#       [6, 6, 6, 6, 6],
#       [7, 7, 7, 7, 7]],
#      [[5, 5, 5, 5, 5],
#       [6, 6, 6, 6, 6],
#       [7, 7, 7, 7, 7]]]), Tensor(shape=[4, 3, 5], dtype=Int32, value=
#     [[[8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2]],
#      [[8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2]],
#      [[8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2]],
#      [[8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2],
#       [8, 9, 0, 1, 2]]]))