Pytorch学习(一)基础语法篇

how to use pytorch

1.Tensor

we can create a tensor just like creating a matrix the default type of a tensor is float

import torch as t
a = t.Tensor([[1,2],[3,4],[5,6]])
a

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

we can also change the datatype of a tensor

b = t.LongTensor([[1,2],[3,4],[5,6]])
b

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

we can also create a tensor filled with zero or random values

c = t.zeros((3,2))
d = t.randn((3,2))
print(c)
print(d)

tensor([[0., 0.],
        [0., 0.],
        [0., 0.]])
tensor([[ 1.2880, -0.1640],
        [-0.2654,  0.7187],
        [-0.3156,  0.4489]])

we can change the value in a tensor we've created

a[0,1] = 100
a

tensor([[  1., 100.],
        [  3.,   4.],
        [  5.,   6.]])

numpy and tensor can transfer from each other

import numpy as np
e = np.array([[1,2],[3,4],[5,6]])
torch_e = t.from_numpy(e)
torch_e

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

2.Variable

Variable consists of data, grad, and grad_fn

data为Tensor中的数值

grad是反向传播梯度

grad_fn是得到该Variable的操作 例如加减乘除

from torch.autograd import Variable
x = Variable(t.Tensor([1]),requires_grad = True)
w = Variable(t.Tensor([2]),requires_grad = True)
b = Variable(t.Tensor([3]),requires_grad = True)
y = w*x+b
y.backward()
print(x.grad)
print(w.grad)
print(b.grad)

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

we can also calculate the grad of a matrix

x = t.randn(3)
x = Variable(x,requires_grad=True)
y = x*2
print(y)
y.backward(t.FloatTensor([1,1,1]))
print(x.grad)

tensor([-2.4801,  0.6291, -0.4250], grad_fn=<MulBackward>)
tensor([2., 2., 2.])

3.dataset

you can define the function len and getitem to write your own dataset

import pandas as pd
from torch.utils.data import Dataset
class myDataset(Dataset):
    def __init__(self, csv_file, txt_file, root_dir, other_file):
        self.csv_data = pd.read_csv(csv_file)
        with open(txt_file, 'r') as f:
            data_list = f.readlines()
        self.txt_data = data_list
        self.root_dir = root_dir
    def __len__(self):
        return len(self.csv_data)
    def __getitem(self,idx):
        data = (self.csv_data[idx],self.txt_data[idx])
        return data

4.nn.Module

from torch import nn
class net_name(nn.Module):
    def __init(self,other_arguments):
        super(net_name, self).__init__()
    def forward(self,x):
        x = self.convl(x)
        return x

5.Optim

1.一阶优化算法

常见的是梯度下降法\(\theta = \theta-\eta\times \frac{\partial J(\theta)}{\partial\theta}\)

2.二阶优化算法

Hessian法