|
np.random.rand（3,5）
返回0-1之间的随机分布数据
np.random.randn（3,5）
返回正态分布的数据
numpy数据和tensor数据之间的相互转换，tensor数据可以和numpy进行相同的操作，还能在GPU上进行加速，所以运行速度较快。
numpy_tensor = np.random.randn(10, 2)
pytorch_tensor1 = torch.Tensor(numpy_tensor)
pytorch_tensor2 = torch.from_numpy(numpy_tensor)
type 返回的是数据类型，列表，数组，tensor
dtype返回的是数据类型，int float
print(type(pytorch_tensor1))
<class ‘torch.Tensor’>
print(pytorch_tensor1.type())
torch.FloatTensor
print(pytorch_tensor1.shape)
torch.Size([10,2])
print(pytorch_tensor1.size())
torch.Size([10,2])
print(pytorch_tensor1.dim())
2
print(pytorch_tensor1.numel())
20
print(type(pytorch_tensor2))
numpy_array = pytorch_tensor1.numpy()
print(type(numpy_array))
<class ‘numpy.ndarray’>
print(numpy_array.dtype)
float32
x = torch.randn(3, 4)
print(x)
print(x.shape)
torch.Size([3, 4])
x = x.unsqueeze(0)
print(x.shape)
torch.Size([1, 3, 4]
x = x.squeeze(0)
print(x.shape)
torch.Size([3, 4])
#squeeze 数据维度的减少 unssqueeze数据维度的增加
x = torch.randn(3, 4, 5)
print(x.shape)
torch.Size([3, 4, 5])
x = x.permute(1, 0, 2)
print(x.shape)
torch.Size([4, 3, 5])
x = x.transpose(0, 2)
print(x.shape)
torch.Size([5, 3, 4])
print(x)
#同一数据的不同维度进行数据交换permute
#仅仅两个维度进行相互交换transpose
Variable
import numpy as np
import torch
from torch.autograd import Variable
x = Variable(torch.Tensor([2]), requires_grad=True)
y = x + 2
z = y**2 + 3
print(z) #简单梯度求导
z.backward() 对变量进行求导
print(x.grad) 提取对x求导的结果
m = Variable(torch.FloatTensor([[2, 3]]), requires_grad=True)
n = Variable(torch.zeros(1, 2))
print(m)
print(n)
print(m.shape)
n[0, 0] = m[0, 0] ** 2
n[0, 1] = m[0, 1] ** 3 #二维的求导
n.backward(torch.ones_like(n)) #此处相当于对每个公式前面添加一个变量
print(m.grad) #对m进行求导
x = Variable(torch.FloatTensor([[3]]), requires_grad=True)
y = x * 2 + x ** 2 + 3
print(y)
18
y.backward(retain_graph=True) #进行求导并保留求导之后的结果
print(x.grad)
8
y.backward() #两次求导的结果进行相加 为 8 +8 =16
print(x.grad)
16
x = Variable(torch.FloatTensor([2, 3]), requires_grad=True)
k = Variable(torch.zeros(2))
k[0] = x[0] ** 2 + 3 * x[1]
k[1] = x[1] ** 2 + 2 * x[0] #矩阵求导 两个公式分别读 x y求导，一共有四个值。k.backward（）中添加的参数就是表示这两个公式分别对x y求导的值是否相加再一起 一下的表达方式可以得出四个值，若在参数中使用tensor.ones_like(n) 则可求出两个相加的值。
print(k)
j = torch.zeros(2, 2)
k.backward(torch.FloatTensor([1, 0]), retain_graph=True)
j[0] = x.grad.data
print(x.grad.data)
x.grad.data.zero_()
print(x.grad.data)
k.backward(torch.FloatTensor([0, 1]))
j[1] = x.grad.data
print(j)
print(x.grad.data)
|