Pytorch入门（一）基本数据类型

本文主要介绍pytoch和numpy的数据之间的转化和pytorch中张量tensor的使用

使用numpy新建一个一维数组

import torch
import numpy as np

np_data = np.arange(6)
print(np_data,type(np_data))

[0 1 2 3 4 5] <class 'numpy.ndarray'>

把数组转化为二维

import torch
import numpy as np

np_data = np.arange(6)
print(np_data,type(np_data))
np_data = np_data.reshape((2,3))
print(np_data,type(np_data))

[0 1 2 3 4 5] <class 'numpy.ndarray'>
[[0 1 2]
 [3 4 5]] <class 'numpy.ndarray'>

这个时候可以看到，数组变为二维的了，两行三列

把二维数组转化为tensor

import torch
import numpy as np

np_data = np.arange(6).reshape((2,3))
torch_data = torch.from_numpy(np_data)
print(
    '\nnumpy:', np_data,
    '\ntorch:', torch_data,
)

numpy: [[0 1 2]
 [3 4 5]] 
torch: tensor([[0, 1, 2],
        [3, 4, 5]]) 

把tensor转为数组

import torch
import numpy as np

np_data = np.arange(6).reshape((2,3))
torch_data = torch.from_numpy(np_data)
tensor2array = torch_data.numpy()
print(
    '\nnumpy:', np_data,
    '\ntorch:', torch_data,
    '\ntensor2array', tensor2array,
)

numpy: [[0 1 2]
 [3 4 5]] 
torch: tensor([[0, 1, 2],
        [3, 4, 5]]) 
tensor2array [[0 1 2]
 [3 4 5]]

随机创建一个二维tensor

import torch

torch_data = torch.randn(2,3)
print(
    '\ntorch_data:',torch_data
)

torch_data: tensor([[ 0.1533, -0.1010,  1.8385],
        [-1.0455,  0.3707,  0.0191]])

构造初始化为0，1的tensor

import torch

torch_data_0 = torch.zeros(2,3)
torch_data_1 = torch.ones(2,3)
print(
    '\ntorch_data_0:',torch_data_0,
    '\ntorch_data_1:',torch_data_1,
)

torch_data_0: tensor([[0., 0., 0.],
        [0., 0., 0.]]) 
torch_data_1: tensor([[1., 1., 1.],
        [1., 1., 1.]])

从python列表直接构造tensor

import torch

torch_data_from_list = torch.Tensor([[1,2,3],[4,5,6]])
print(
    '\ntorch_data_from_list:',torch_data_from_list,
)

torch_data_from_list: tensor([[1., 2., 3.],
        [4., 5., 6.]])

需要注意的是，在pytorch中，默认的数据类型就是torch.FloatTensor，所以在这里torch.FloatTensor就等于torch.Tensor

import torch

torch_data_from_list = torch.Tensor([[1,2,3],[4,5,6]])
torch_data_from_list_32 = torch.FloatTensor([[1,2,3],[4,5,6]])
print(
    '\ntorch_data_from_list:',torch_data_from_list,
    '\ntorch_data_from_list_32',torch_data_from_list_32
)

torch_data_from_list: tensor([[1., 2., 3.],
        [4., 5., 6.]]) 
torch_data_from_list_32 tensor([[1., 2., 3.],
        [4., 5., 6.]])