深度学习入门笔记（5）——什么是tensor？

深度学习入门笔记（5）——tensor

（1）定位：操作对象
（2）概念：张量，表示一个矩阵，与numpy中的ndarray对应。
（3）与numpy数组区别：tensor可以在GPU上运行，ndarray只能在CPU上运行。
（4）数据类型：int、float等等都可以，默认情况为float
（5）声明：

# 定义两行两列列给定数值的矩阵
matrix = torch.Tensor([[1, 1], [2, 2]])
# 打印矩阵数据类型、大小及样式
print("the matrix type is {}".format(type(matrix)))
print("the matrix size is {}".format(matrix.shape))
print("the matrix is \n{}".format(matrix))

运行结果

the matrix type is <class 'torch.Tensor'>
the matrix size is torch.Size([2, 2])
the matrix is 
tensor([[1., 1.],
        [2., 2.]])
        
Process finished with exit code 0

（6）zeros和ones：

# 定义矩阵
matrix_zero = torch.zeros((2,2))
matrix_one = torch.ones((2,2))
# 打印矩阵数据类型、大小及样式
print("the matrix type is {}".format(type(matrix_zero)))
print("the matrix size is {}".format(matrix_zero.shape))
print("the matrix is \n{}".format(matrix_zero))

print("the matrix type is {}".format(type(matrix_one)))
print("the matrix size is {}".format(matrix_one.shape))
print("the matrix is \n{}".format(matrix_one))

运行结果

the matrix type is <class 'torch.Tensor'>
the matrix size is torch.Size([2, 2])
the matrix is 
tensor([[0., 0.],
        [0., 0.]])
the matrix type is <class 'torch.Tensor'>
the matrix size is torch.Size([2, 2])
the matrix is 
tensor([[1., 1.],
        [1., 1.]])

Process finished with exit code 0

这里我们也可以发现，tensor默认为float
（7）索引

# 定义两行两列列给定数值的矩阵
matrix = torch.Tensor([[1,1],[2,2]])
# 索引
print("the last value of this matrix is {}".format(matrix[1,1]))
# 改变最后一个元素的值
matrix[1,1] = 6
print("after changed,the last value of this matrix is {}".format(matrix[1,1]))

运行结果

the last value of this matrix is 2.0
after changed,the last value of this matrix is 6.0

Process finished with exit code 0

（8）与ndarray格式转换

# 定义torch和numpy格式下的矩阵（数组）
matrix = torch.Tensor([[1,1],[2,2]])
array = np.array([[2,2],[1,1]])
# array转换为torch格式
print("before change , the array's type is {}".format(type(array)))
matrix_change = torch.from_numpy(array)
print("before change , the array's type is {}".format(type(matrix_change)))

运行结果

before change , the array's type is <class 'numpy.ndarray'>
before change , the array's type is <class 'torch.Tensor'>

Process finished with exit code 0

最后

欢迎大家在评论区留言讨论！