201_PyTorch中文教程：Torch与Numpy互操作

Numpy是经典的数学计算库，Torch中的Tensor可以与之互相转换，从而可以充分利用二者的计算函数和模型，以及使用其它支持Numpy的软件库和工具。但需注意，转换需要花费额外的内存和CPU等计算资源。

• 更多参考: https://morvanzhou.github.io/tutorials/
• 油管频道：https://www.youtube.com/user/MorvanZhou

依赖软件包:

• torch
• numpy

Torch的更多数学操作，参考： http://pytorch.org/docs/torch.html#math-operations

import torch
import numpy as np

# 转换 numpy 为 tensor，或者转回来。
np_data = np.arange(6).reshape((2, 3))
torch_data = torch.from_numpy(np_data)
tensor2array = torch_data.numpy()
print(
    '\nnumpy array:', np_data,          # [[0 1 2], [3 4 5]]
    '\ntorch tensor:', torch_data,      #  0  1  2 \n 3  4  5    [torch.LongTensor of size 2x3]
    '\ntensor to array:', tensor2array, # [[0 1 2], [3 4 5]]
)

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

# 求绝对值
data = [-1, -2, 1, 2]
tensor = torch.FloatTensor(data)  # 32-bit floating point
print(
    '\nabs',
    '\nnumpy: ', np.abs(data),          # [1 2 1 2]
    '\ntorch: ', torch.abs(tensor)      # [1 2 1 2]
)

abs 
numpy:  [1 2 1 2] 
torch:  tensor([1., 2., 1., 2.])

tensor.abs()

tensor([1., 2., 1., 2.])

# 求sin值
print(
    '\nsin',
    '\nnumpy: ', np.sin(data),      # [-0.84147098 -0.90929743  0.84147098  0.90929743]
    '\ntorch: ', torch.sin(tensor)  # [-0.8415 -0.9093  0.8415  0.9093]
)

sin 
numpy:  [-0.84147098 -0.90929743  0.84147098  0.90929743] 
torch:  tensor([-0.8415, -0.9093,  0.8415,  0.9093])

tensor.sigmoid()

tensor([0.2689, 0.1192, 0.7311, 0.8808])

tensor.exp()

tensor([0.3679, 0.1353, 2.7183, 7.3891])

# mean
print(
    '\nmean',
    '\nnumpy: ', np.mean(data),         # 0.0
    '\ntorch: ', torch.mean(tensor)     # 0.0
)

mean 
numpy:  0.0 
torch:  tensor(0.)

# 矩阵乘法，matrix multiplication
data = [[1,2], [3,4]]
tensor = torch.FloatTensor(data)  # 32-bit floating point
# correct method
print(
    '\nmatrix multiplication (matmul)',
    '\nnumpy: ', np.matmul(data, data),     # [[7, 10], [15, 22]]
    '\ntorch: ', torch.mm(tensor, tensor)   # [[7, 10], [15, 22]]
)

matrix multiplication (matmul) 
numpy:  [[ 7 10]
 [15 22]] 
torch:  tensor([[ 7., 10.],
        [15., 22.]])

# 不正确的方法
data = np.array(data)
tensor = torch.Tensor(data)

# 参考：https://www.cnblogs.com/yangzhaonan/p/10439416.html
print(
    '\nmatrix multiplication (dot)',
    '\nnumpy: ', data.dot(data),        # [[7, 10], [15, 22]]
    '\ntorch: ', torch.dot(tensor.dot(tensor))     # NOT WORKING! Beware that torch.dot does not broadcast, only works for 1-dimensional tensor
)

---------------------------------------------------------------------------

TypeError                                 Traceback (most recent call last)

<ipython-input-22-de97c709d870> in <module>()
      7     '\nmatrix multiplication (dot)',
      8     '\nnumpy: ', data.dot(data),        # [[7, 10], [15, 22]]
----> 9     '\ntorch: ', torch.dot(tensor.dot(tensor))     # NOT WORKING! Beware that torch.dot does not broadcast, only works for 1-dimensional tensor
     10 )


TypeError: dot() missing 1 required positional arguments: "tensor"

Note that:

torch.dot(tensor1, tensor2) → float

Computes the dot product (inner product) of two tensors. Both tensors are treated as 1-D vectors.

tensor.mm(tensor)

tensor([[ 7., 10.],
        [15., 22.]])

tensor * tensor

tensor([[ 1.,  4.],
        [ 9., 16.]])

torch.dot(torch.Tensor([2, 3]), torch.Tensor([2, 1]))

tensor(7.)